feat(#707): secure Discord service ingress

.prettierignore

@@ -5,3 +5,5 @@ pnpm-lock.yaml
 **/drizzle
 **/.next
 .claude/
+docs/tess/TASKS.md
+docs/scratchpads/

apps/gateway/src/chat/chat.gateway-auth.ts

@@ -1,3 +1,4 @@
+import { timingSafeEqual } from 'node:crypto';
 import type { IncomingHttpHeaders } from 'node:http';
 import { fromNodeHeaders } from 'better-auth/node';
 
@@ -12,6 +13,19 @@ export interface SessionAuth {
   };
 }
 
+export function validateDiscordServiceToken(
+  candidate: unknown,
+  expected: string | undefined,
+): boolean {
+  if (typeof candidate !== 'string' || !expected) return false;
+  const candidateBuffer = Buffer.from(candidate);
+  const expectedBuffer = Buffer.from(expected);
+  return (
+    candidateBuffer.length === expectedBuffer.length &&
+    timingSafeEqual(candidateBuffer, expectedBuffer)
+  );
+}
+
 export async function validateSocketSession(
   headers: IncomingHttpHeaders,
   auth: SessionAuth,

apps/gateway/src/chat/chat.gateway.ts

@@ -11,6 +11,11 @@ import {
 } from '@nestjs/websockets';
 import { Server, Socket } from 'socket.io';
 import type { AgentSessionEvent } from '@mariozechner/pi-coding-agent';
+import {
+  verifyDiscordIngressEnvelope,
+  type DiscordIngressEnvelope,
+  type DiscordIngressPayload,
+} from '@mosaicstack/discord-plugin';
 import type { Auth } from '@mosaicstack/auth';
 import type { Brain } from '@mosaicstack/brain';
 import type {
@@ -28,7 +33,8 @@ import { CommandExecutorService } from '../commands/command-executor.service.js'
 import { RoutingEngineService } from '../agent/routing/routing-engine.service.js';
 import { v4 as uuid } from 'uuid';
 import { ChatSocketMessageDto } from './chat.dto.js';
-import { validateSocketSession } from './chat.gateway-auth.js';
+import { validateDiscordServiceToken, validateSocketSession } from './chat.gateway-auth.js';
+import { DiscordReplayProtector } from '../plugin/discord-replay-protector.js';
 
 /** Per-client state tracking streaming accumulation for persistence. */
 interface ClientSession {
@@ -50,6 +56,37 @@ interface ClientSession {
  */
 const modelOverrides = new Map<string, string>();
 
+function isDiscordIngressEnvelope(value: unknown): value is DiscordIngressEnvelope {
+  if (typeof value !== 'object' || value === null) return false;
+  const envelope = value as { payload?: unknown; signature?: unknown };
+  if (
+    typeof envelope.signature !== 'string' ||
+    typeof envelope.payload !== 'object' ||
+    envelope.payload === null
+  ) {
+    return false;
+  }
+  const payload = envelope.payload as Record<string, unknown>;
+  return [
+    payload['correlationId'],
+    payload['messageId'],
+    payload['guildId'],
+    payload['channelId'],
+    payload['userId'],
+    payload['conversationId'],
+    payload['content'],
+  ].every((field: unknown): boolean => typeof field === 'string');
+}
+
+function isChatSocketMessage(value: unknown): value is ChatSocketMessageDto {
+  if (typeof value !== 'object' || value === null) return false;
+  const payload = value as { content?: unknown; conversationId?: unknown };
+  return (
+    typeof payload.content === 'string' &&
+    (payload.conversationId === undefined || typeof payload.conversationId === 'string')
+  );
+}
+
 @WebSocketGateway({
   cors: {
     origin: process.env['GATEWAY_CORS_ORIGIN'] ?? 'http://localhost:3000',
@@ -62,6 +99,7 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
 
   private readonly logger = new Logger(ChatGateway.name);
   private readonly clientSessions = new Map<string, ClientSession>();
+  private readonly discordReplayProtector = new DiscordReplayProtector();
 
   constructor(
     @Inject(AgentService) private readonly agentService: AgentService,
@@ -77,6 +115,13 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
   }
 
   async handleConnection(client: Socket): Promise<void> {
+    const serviceToken = client.handshake.auth['discordServiceToken'];
+    if (validateDiscordServiceToken(serviceToken, process.env['DISCORD_SERVICE_TOKEN'])) {
+      client.data.discordService = true;
+      this.logger.log(`Authenticated Discord service connected: ${client.id}`);
+      return;
+    }
+
     const session = await validateSocketSession(client.handshake.headers, this.auth);
     if (!session) {
       this.logger.warn(`Rejected unauthenticated WebSocket client: ${client.id}`);
@@ -87,8 +132,6 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
     client.data.user = session.user;
     client.data.session = session.session;
     this.logger.log(`Client connected: ${client.id}`);
-
-    // Broadcast command manifest to the newly connected client
     client.emit('commands:manifest', { manifest: this.commandRegistry.getManifest() });
   }
 
@@ -105,12 +148,41 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
   @SubscribeMessage('message')
   async handleMessage(
     @ConnectedSocket() client: Socket,
-    @MessageBody() data: ChatSocketMessageDto,
+    @MessageBody() rawData: unknown,
   ): Promise<void> {
+    let discordIngress: DiscordIngressPayload | null = null;
+    let data: ChatSocketMessageDto;
+    if (client.data.discordService) {
+      if (!isDiscordIngressEnvelope(rawData)) {
+        this.logger.warn(`Rejected malformed Discord ingress from ${client.id}`);
+        return;
+      }
+      discordIngress = this.resolveDiscordIngress(client, rawData);
+      if (!discordIngress) return;
+      data = { conversationId: discordIngress.conversationId, content: discordIngress.content };
+    } else {
+      if (!isChatSocketMessage(rawData)) {
+        this.logger.warn(`Rejected malformed chat message from ${client.id}`);
+        return;
+      }
+      data = rawData;
+    }
     const conversationId = data.conversationId ?? uuid();
-    const userId = (client.data.user as { id: string } | undefined)?.id;
+    const discordServiceUserId = process.env['DISCORD_SERVICE_USER_ID'];
+    if (discordIngress && !discordServiceUserId) {
+      this.logger.warn(
+        `Rejected Discord ingress without configured service owner from ${client.id}`,
+      );
+      return;
+    }
+    const userId = discordIngress
+      ? discordServiceUserId
+      : (client.data.user as { id: string } | undefined)?.id;
+    const correlationId = discordIngress?.correlationId;
 
-    this.logger.log(`Message from ${client.id} in conversation ${conversationId}`);
+    this.logger.log(
+      `Message from ${client.id} in conversation ${conversationId}${correlationId ? ` correlation=${correlationId}` : ''}`,
+    );
 
     // Ensure agent session exists for this conversation
     let sessionRoutingDecision: RoutingDecisionInfo | undefined;
@@ -213,6 +285,13 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
             content: data.content,
             metadata: {
               timestamp: new Date().toISOString(),
+              ...(correlationId
+                ? {
+                    correlationId,
+                    discordMessageId: discordIngress?.messageId,
+                    discordUserId: discordIngress?.userId,
+                  }
+                : {}),
             },
           },
           userId,
@@ -271,7 +350,17 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
     }
 
     // Send acknowledgment
-    client.emit('message:ack', { conversationId, messageId: uuid() });
+    client.emit('message:ack', {
+      conversationId,
+      messageId: uuid(),
+      ...(correlationId
+        ? {
+            correlationId,
+            discordMessageId: discordIngress?.messageId,
+            discordUserId: discordIngress?.userId,
+          }
+        : {}),
+    });
 
     // Dispatch to agent
     try {
@@ -442,6 +531,39 @@ export class ChatGateway implements OnGatewayInit, OnGatewayConnection, OnGatewa
    * Creates it if absent — safe to call concurrently since a duplicate insert
    * would fail on the PK constraint and be caught here.
    */
+  private resolveDiscordIngress(
+    client: Socket,
+    envelope: DiscordIngressEnvelope,
+  ): DiscordIngressPayload | null {
+    const payload = verifyDiscordIngressEnvelope(
+      envelope,
+      process.env['DISCORD_SERVICE_TOKEN'] ?? '',
+      {
+        guildIds: this.readDiscordAllowlist('DISCORD_ALLOWED_GUILD_IDS'),
+        channelIds: this.readDiscordAllowlist('DISCORD_ALLOWED_CHANNEL_IDS'),
+        userIds: this.readDiscordAllowlist('DISCORD_ALLOWED_USER_IDS'),
+      },
+    );
+    if (!payload) {
+      this.logger.warn(`Rejected invalid Discord ingress envelope from ${client.id}`);
+      return null;
+    }
+    if (!this.discordReplayProtector.claim(payload.messageId)) {
+      this.logger.warn(
+        `Rejected replayed Discord message=${payload.messageId} correlation=${payload.correlationId}`,
+      );
+      return null;
+    }
+    return payload;
+  }
+
+  private readDiscordAllowlist(name: string): string[] {
+    return (process.env[name] ?? '')
+      .split(',')
+      .map((id: string): string => id.trim())
+      .filter((id: string): boolean => id.length > 0);
+  }
+
   private async ensureConversation(conversationId: string, userId: string): Promise<void> {
     try {
       const existing = await this.brain.conversations.findById(conversationId, userId);

apps/gateway/src/federation/client/__tests__/query-source.service.spec.ts

@@ -0,0 +1,255 @@
+import 'reflect-metadata';
+import { describe, expect, it, vi } from 'vitest';
+import type { Db } from '@mosaicstack/db';
+import type { FederationListResponse } from '@mosaicstack/types';
+import {
+  FederationClientError,
+  type FederationClientService,
+} from '../federation-client.service.js';
+import { type QuerySourceError, QuerySourceService } from '../query-source.service.js';
+
+interface TestRow {
+  id: string;
+  title: string;
+}
+
+interface PeerRow {
+  id: string;
+  commonName: string;
+  endpointUrl: string | null;
+  clientKeyPem: string | null;
+  state: 'active' | 'pending' | 'suspended' | 'revoked';
+}
+
+const LOCAL_ROWS: TestRow[] = [
+  { id: 'local-1', title: 'Local One' },
+  { id: 'local-2', title: 'Local Two' },
+];
+
+const PEER_A: PeerRow = {
+  id: 'aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa',
+  commonName: 'peer-a',
+  endpointUrl: 'https://peer-a.example.com',
+  clientKeyPem: 'sealed-key-a',
+  state: 'active',
+};
+
+const PEER_B: PeerRow = {
+  id: 'bbbbbbbb-bbbb-bbbb-bbbb-bbbbbbbbbbbb',
+  commonName: 'peer-b',
+  endpointUrl: 'https://peer-b.example.com',
+  clientKeyPem: 'sealed-key-b',
+  state: 'active',
+};
+
+const PEER_LOCALHOST: PeerRow = {
+  id: 'cccccccc-cccc-cccc-cccc-cccccccccccc',
+  commonName: 'peer-localhost',
+  endpointUrl: 'https://localhost:3001',
+  clientKeyPem: 'sealed-key-c',
+  state: 'active',
+};
+
+function makeDb(activePeers: PeerRow[]): Db {
+  const orderBy = vi.fn().mockResolvedValue(activePeers);
+  const where = vi.fn().mockReturnValue({ orderBy });
+  const from = vi.fn().mockReturnValue({ where });
+  const select = vi.fn().mockReturnValue({ from });
+
+  return {
+    select,
+    insert: vi.fn(),
+    update: vi.fn(),
+    delete: vi.fn(),
+    transaction: vi.fn(),
+  } as unknown as Db;
+}
+
+function makeFederationClient(
+  list: (
+    peerId: string,
+    resource: string,
+    request: Record<string, unknown>,
+  ) => Promise<FederationListResponse<TestRow>>,
+): FederationClientService {
+  return {
+    list: list as unknown as FederationClientService['list'],
+  } as FederationClientService;
+}
+
+function makeLocalResponse(rows: TestRow[] = LOCAL_ROWS): Promise<FederationListResponse<TestRow>> {
+  return Promise.resolve({ items: rows });
+}
+
+describe('QuerySourceService', () => {
+  it('routes source="local" to the local executor and tags rows as local', async () => {
+    const list = vi.fn(async (): Promise<FederationListResponse<TestRow>> => ({ items: [] }));
+    const service = new QuerySourceService(makeDb([PEER_A]), makeFederationClient(list));
+
+    const result = await service.list<TestRow>({
+      source: 'local',
+      resource: 'tasks',
+      request: { cursor: 'ignored-for-local-test' },
+      local: () => makeLocalResponse(),
+    });
+
+    expect(result).toEqual({
+      items: [
+        { id: 'local-1', title: 'Local One', _source: 'local' },
+        { id: 'local-2', title: 'Local Two', _source: 'local' },
+      ],
+    });
+    expect(list).not.toHaveBeenCalled();
+  });
+
+  it('routes source="federated:<host>" to the matching active peer and tags rows with peer commonName', async () => {
+    const list = vi.fn(
+      async (): Promise<FederationListResponse<TestRow>> => ({
+        items: [{ id: 'remote-1', title: 'Remote One' }],
+      }),
+    );
+    const service = new QuerySourceService(makeDb([PEER_A, PEER_B]), makeFederationClient(list));
+
+    const result = await service.list<TestRow>({
+      source: 'federated:peer-b.example.com',
+      resource: 'tasks',
+      request: { status: 'open' },
+      local: () => makeLocalResponse(),
+    });
+
+    expect(result).toEqual({
+      items: [{ id: 'remote-1', title: 'Remote One', _source: 'peer-b' }],
+    });
+    expect(list).toHaveBeenCalledWith(PEER_B.id, 'tasks', { status: 'open' });
+  });
+
+  it('matches federated hosts by endpoint host including non-default port', async () => {
+    const list = vi.fn(
+      async (): Promise<FederationListResponse<TestRow>> => ({
+        items: [{ id: 'remote-port', title: 'Remote Port' }],
+      }),
+    );
+    const service = new QuerySourceService(makeDb([PEER_LOCALHOST]), makeFederationClient(list));
+
+    const result = await service.list<TestRow>({
+      source: 'federated:localhost:3001',
+      resource: 'tasks',
+      request: {},
+      local: () => makeLocalResponse(),
+    });
+
+    expect(result).toEqual({
+      items: [{ id: 'remote-port', title: 'Remote Port', _source: 'peer-localhost' }],
+    });
+    expect(list).toHaveBeenCalledWith(PEER_LOCALHOST.id, 'tasks', {});
+  });
+
+  it('fans out source="all" to local plus every active outbound peer in parallel and merges tagged rows', async () => {
+    const callOrder: string[] = [];
+    const list = vi.fn(async (peerId: string): Promise<FederationListResponse<TestRow>> => {
+      callOrder.push(`remote-start:${peerId}`);
+      await Promise.resolve();
+      return {
+        items: [{ id: `remote-${peerId.slice(0, 1)}`, title: `Remote ${peerId.slice(0, 1)}` }],
+      };
+    });
+    const service = new QuerySourceService(makeDb([PEER_A, PEER_B]), makeFederationClient(list));
+
+    const result = await service.list<TestRow>({
+      source: 'all',
+      resource: 'tasks',
+      request: { limit: 25 },
+      local: async () => {
+        callOrder.push('local-start');
+        await Promise.resolve();
+        return { items: [{ id: 'local-1', title: 'Local One' }] };
+      },
+    });
+
+    expect(result).toEqual({
+      items: [
+        { id: 'local-1', title: 'Local One', _source: 'local' },
+        { id: 'remote-a', title: 'Remote a', _source: 'peer-a' },
+        { id: 'remote-b', title: 'Remote b', _source: 'peer-b' },
+      ],
+    });
+    expect(list).toHaveBeenCalledTimes(2);
+    expect(callOrder).toEqual([
+      'local-start',
+      `remote-start:${PEER_A.id}`,
+      `remote-start:${PEER_B.id}`,
+    ]);
+  });
+
+  it('marks source="all" as partial and truncated when any subquery returns a cursor', async () => {
+    const list = vi.fn(
+      async (): Promise<FederationListResponse<TestRow>> => ({
+        items: [{ id: 'remote-a', title: 'Remote A' }],
+        nextCursor: 'remote-next',
+      }),
+    );
+    const service = new QuerySourceService(makeDb([PEER_A]), makeFederationClient(list));
+
+    const result = await service.list<TestRow>({
+      source: 'all',
+      resource: 'tasks',
+      request: {},
+      local: () => makeLocalResponse([{ id: 'local-1', title: 'Local One' }]),
+    });
+
+    expect(result).toEqual({
+      items: [
+        { id: 'local-1', title: 'Local One', _source: 'local' },
+        { id: 'remote-a', title: 'Remote A', _source: 'peer-a' },
+      ],
+      _partial: true,
+      _truncated: true,
+    });
+  });
+
+  it('returns _partial=true for source="all" when one peer fails without dropping successful sources', async () => {
+    const list = vi.fn(async (peerId: string): Promise<FederationListResponse<TestRow>> => {
+      if (peerId === PEER_B.id) {
+        throw new FederationClientError({
+          code: 'NETWORK',
+          message: 'peer unavailable',
+          peerId,
+        });
+      }
+      return { items: [{ id: 'remote-a', title: 'Remote A' }] };
+    });
+    const service = new QuerySourceService(makeDb([PEER_A, PEER_B]), makeFederationClient(list));
+
+    const result = await service.list<TestRow>({
+      source: 'all',
+      resource: 'tasks',
+      request: {},
+      local: () => makeLocalResponse([{ id: 'local-1', title: 'Local One' }]),
+    });
+
+    expect(result).toEqual({
+      items: [
+        { id: 'local-1', title: 'Local One', _source: 'local' },
+        { id: 'remote-a', title: 'Remote A', _source: 'peer-a' },
+      ],
+      _partial: true,
+    });
+  });
+
+  it('throws QuerySourceError when a federated host does not match an active outbound peer', async () => {
+    const list = vi.fn(async (): Promise<FederationListResponse<TestRow>> => ({ items: [] }));
+    const service = new QuerySourceService(makeDb([PEER_A]), makeFederationClient(list));
+
+    await expect(
+      service.list<TestRow>({
+        source: 'federated:missing.example.com',
+        resource: 'tasks',
+        request: {},
+        local: () => makeLocalResponse(),
+      }),
+    ).rejects.toMatchObject({
+      name: 'QuerySourceError',
+      code: 'PEER_NOT_FOUND',
+    } satisfies Partial<QuerySourceError>);
+  });
+});

apps/gateway/src/federation/client/index.ts

@@ -11,3 +11,13 @@ export {
   type FederationClientErrorCode,
   type FederationClientErrorOptions,
 } from './federation-client.service.js';
+export {
+  QuerySourceService,
+  QuerySourceError,
+  type QuerySource,
+  type QuerySourceErrorCode,
+  type QuerySourceErrorOptions,
+  type QuerySourceListOptions,
+  type QuerySourceListResponse,
+  type LocalListExecutor,
+} from './query-source.service.js';

apps/gateway/src/federation/client/query-source.service.ts

@@ -0,0 +1,261 @@
+/**
+ * QuerySourceService — gateway query source router (FED-M3-09).
+ *
+ * Accepts the federation query-layer `source` selector and routes list-style
+ * reads to local storage, one federated peer, or all active outbound peers.
+ *
+ * `source: "all"` is intentionally tolerant of per-peer failures: local data
+ * and successful peer responses are returned, and the envelope is marked
+ * `_partial: true`.  Local failures still reject because there is no safe local
+ * fallback and the gateway's own storage is expected to be authoritative.
+ */
+
+import { Inject, Injectable, Logger } from '@nestjs/common';
+import { and, eq, federationPeers, isNotNull, type Db } from '@mosaicstack/db';
+import {
+  SOURCE_LOCAL,
+  tagWithSource,
+  type FederationListResponse,
+  type SourceTag,
+} from '@mosaicstack/types';
+import { DB } from '../../database/database.module.js';
+import { FederationClientService } from './federation-client.service.js';
+
+export type QuerySource = 'local' | 'all' | `federated:${string}`;
+
+export type QuerySourceErrorCode = 'INVALID_SOURCE' | 'PEER_NOT_FOUND';
+
+export interface QuerySourceErrorOptions {
+  code: QuerySourceErrorCode;
+  message: string;
+  source: string;
+}
+
+export class QuerySourceError extends Error {
+  readonly code: QuerySourceErrorCode;
+  readonly source: string;
+
+  constructor(opts: QuerySourceErrorOptions) {
+    super(opts.message);
+    this.name = 'QuerySourceError';
+    this.code = opts.code;
+    this.source = opts.source;
+  }
+}
+
+export type LocalListExecutor<T extends object> = () => Promise<FederationListResponse<T> | T[]>;
+
+export interface QuerySourceListOptions<T extends object> {
+  source: QuerySource;
+  resource: string;
+  request?: Record<string, unknown>;
+  local: LocalListExecutor<T>;
+}
+
+export type QuerySourceListResponse<T extends object> = FederationListResponse<T & SourceTag>;
+
+interface OutboundPeer {
+  id: string;
+  commonName: string;
+  endpointUrl: string;
+}
+
+interface TaggedList<T extends object> {
+  items: Array<T & SourceTag>;
+  partial: boolean;
+  truncated: boolean;
+  nextCursor?: string;
+}
+
+@Injectable()
+export class QuerySourceService {
+  private readonly logger = new Logger(QuerySourceService.name);
+
+  constructor(
+    @Inject(DB) private readonly db: Db,
+    @Inject(FederationClientService) private readonly federationClient: FederationClientService,
+  ) {}
+
+  async list<T extends object>(
+    options: QuerySourceListOptions<T>,
+  ): Promise<QuerySourceListResponse<T>> {
+    const request = options.request ?? {};
+
+    if (options.source === 'local') {
+      const local = await this.runLocal(options.local);
+      return this.toResponse(this.tagList(local, SOURCE_LOCAL));
+    }
+
+    if (options.source === 'all') {
+      return this.listAll(options.resource, request, options.local);
+    }
+
+    if (options.source.startsWith('federated:')) {
+      const host = options.source.slice('federated:'.length).trim();
+      if (!host) {
+        throw new QuerySourceError({
+          code: 'INVALID_SOURCE',
+          message: 'Federated source must include a host after federated:',
+          source: options.source,
+        });
+      }
+
+      const peer = await this.findPeerByHost(host, options.source);
+      const remote = await this.federationClient.list<T>(peer.id, options.resource, request);
+      return this.toResponse(this.tagList(remote, peer.commonName));
+    }
+
+    throw new QuerySourceError({
+      code: 'INVALID_SOURCE',
+      message: `Unsupported query source: ${options.source}`,
+      source: options.source,
+    });
+  }
+
+  private async listAll<T extends object>(
+    resource: string,
+    request: Record<string, unknown>,
+    local: LocalListExecutor<T>,
+  ): Promise<QuerySourceListResponse<T>> {
+    const peers = await this.listActiveOutboundPeers();
+
+    const localPromise = this.runLocal(local).then((response) =>
+      this.tagList(response, SOURCE_LOCAL),
+    );
+    const remotePromises = peers.map(async (peer: OutboundPeer): Promise<TaggedList<T> | null> => {
+      try {
+        const response = await this.federationClient.list<T>(peer.id, resource, request);
+        return this.tagList(response, peer.commonName);
+      } catch (error: unknown) {
+        this.logger.warn(
+          `Federated query to peer ${peer.commonName} (${peer.id}) failed; returning partial all-source response: ${
+            error instanceof Error ? error.message : String(error)
+          }`,
+        );
+        return null;
+      }
+    });
+
+    const [localResult, ...remoteResults] = await Promise.all([localPromise, ...remotePromises]);
+    const successfulRemoteResults = remoteResults.filter(
+      (result: TaggedList<T> | null): result is TaggedList<T> => result !== null,
+    );
+    const allResults = [localResult, ...successfulRemoteResults];
+    const peerFailure = successfulRemoteResults.length !== peers.length;
+
+    return this.mergeTaggedLists(allResults, peerFailure);
+  }
+
+  private async runLocal<T extends object>(
+    local: LocalListExecutor<T>,
+  ): Promise<FederationListResponse<T>> {
+    const response = await local();
+    if (Array.isArray(response)) {
+      return { items: response };
+    }
+    return response;
+  }
+
+  private tagList<T extends object>(
+    response: FederationListResponse<T>,
+    source: string,
+  ): TaggedList<T> {
+    return {
+      items: tagWithSource(response.items, source),
+      partial: response._partial === true,
+      truncated: response._truncated === true || response.nextCursor !== undefined,
+      nextCursor: response.nextCursor,
+    };
+  }
+
+  private mergeTaggedLists<T extends object>(
+    lists: Array<TaggedList<T>>,
+    peerFailure: boolean,
+  ): QuerySourceListResponse<T> {
+    const items = lists.flatMap((list: TaggedList<T>) => list.items);
+    const partial =
+      peerFailure ||
+      lists.some((list: TaggedList<T>) => list.partial || list.nextCursor !== undefined);
+    const truncated = lists.some((list: TaggedList<T>) => list.truncated);
+
+    const response: QuerySourceListResponse<T> = { items };
+    if (partial) {
+      response._partial = true;
+    }
+    if (truncated) {
+      response._truncated = true;
+    }
+    return response;
+  }
+
+  private toResponse<T extends object>(tagged: TaggedList<T>): QuerySourceListResponse<T> {
+    const response: QuerySourceListResponse<T> = {
+      items: tagged.items,
+    };
+    if (tagged.nextCursor !== undefined) {
+      response.nextCursor = tagged.nextCursor;
+    }
+    if (tagged.partial) {
+      response._partial = true;
+    }
+    if (tagged.truncated) {
+      response._truncated = true;
+    }
+    return response;
+  }
+
+  private async findPeerByHost(sourceHost: string, source: string): Promise<OutboundPeer> {
+    const host = normalizeHost(sourceHost);
+    const peers = await this.listActiveOutboundPeers();
+    const peer = peers.find((candidate: OutboundPeer) => {
+      const commonName = normalizeHost(candidate.commonName);
+      const endpointHosts = endpointHostKeys(candidate.endpointUrl).map((endpointHost: string) =>
+        normalizeHost(endpointHost),
+      );
+      return commonName === host || endpointHosts.includes(host);
+    });
+
+    if (!peer) {
+      throw new QuerySourceError({
+        code: 'PEER_NOT_FOUND',
+        message: `No active outbound federation peer matches source ${source}`,
+        source,
+      });
+    }
+
+    return peer;
+  }
+
+  private async listActiveOutboundPeers(): Promise<OutboundPeer[]> {
+    const rows = await this.db
+      .select({
+        id: federationPeers.id,
+        commonName: federationPeers.commonName,
+        endpointUrl: federationPeers.endpointUrl,
+      })
+      .from(federationPeers)
+      .where(
+        and(
+          eq(federationPeers.state, 'active'),
+          isNotNull(federationPeers.endpointUrl),
+          isNotNull(federationPeers.clientKeyPem),
+        ),
+      )
+      .orderBy(federationPeers.commonName);
+
+    return rows.filter((row): row is OutboundPeer => typeof row.endpointUrl === 'string');
+  }
+}
+
+function normalizeHost(host: string): string {
+  return host.trim().toLowerCase();
+}
+
+function endpointHostKeys(endpointUrl: string): string[] {
+  try {
+    const url = new URL(endpointUrl);
+    return Array.from(new Set([url.host, url.hostname].filter((host: string) => host.length > 0)));
+  } catch {
+    return [];
+  }
+}

apps/gateway/src/federation/federation.module.ts

@@ -4,26 +4,35 @@ import { CaService } from './ca.service.js';
 import { EnrollmentController } from './enrollment.controller.js';
 import { EnrollmentService } from './enrollment.service.js';
 import { FederationController } from './federation.controller.js';
+import { CapabilitiesController } from './server/verbs/capabilities.controller.js';
 import { GrantsService } from './grants.service.js';
-import { FederationClientService } from './client/index.js';
-import { FederationAuthGuard } from './server/index.js';
+import { FederationClientService, QuerySourceService } from './client/index.js';
+import { FederationAuthGuard, FederationScopeService } from './server/index.js';
+import { ListController } from './server/verbs/list.controller.js';
+import { FederationListQueryService } from './server/verbs/list-query.service.js';
 
 @Module({
-  controllers: [EnrollmentController, FederationController],
+  controllers: [EnrollmentController, FederationController, CapabilitiesController, ListController],
   providers: [
     AdminGuard,
     CaService,
     EnrollmentService,
     GrantsService,
     FederationClientService,
+    QuerySourceService,
     FederationAuthGuard,
+    FederationScopeService,
+    FederationListQueryService,
   ],
   exports: [
     CaService,
     EnrollmentService,
     GrantsService,
     FederationClientService,
+    QuerySourceService,
     FederationAuthGuard,
+    FederationScopeService,
+    FederationListQueryService,
   ],
 })
 export class FederationModule {}

apps/gateway/src/federation/server/__tests__/scope.service.spec.ts

@@ -0,0 +1,324 @@
+/**
+ * Unit tests for FederationScopeService (FED-M3-04).
+ *
+ * Coverage:
+ *  - resource allowlist deny
+ *  - excluded resource deny
+ *  - invalid scope deny
+ *  - invalid requested limit deny
+ *  - native RBAC deny as subjectUserId
+ *  - scope/native filter intersection for personal and team rows
+ *  - native RBAC personal deny wins over scope include_personal allow/default
+ *  - max_rows_per_query cap
+ */
+
+import { beforeEach, describe, expect, it, vi } from 'vitest';
+import { FederationScopeService, type FederationNativeRbacEvaluator } from '../scope.service.js';
+import type { FederationContext } from '../federation-context.js';
+
+const GRANT_ID = 'grant-1';
+const PEER_ID = 'peer-1';
+const SUBJECT_USER_ID = 'user-1';
+
+function makeContext(scope: Record<string, unknown>): FederationContext {
+  return {
+    grantId: GRANT_ID,
+    peerId: PEER_ID,
+    subjectUserId: SUBJECT_USER_ID,
+    scope,
+  };
+}
+
+function makeNativeRbac(
+  result: Awaited<ReturnType<FederationNativeRbacEvaluator['evaluateReadAccess']>>,
+): FederationNativeRbacEvaluator {
+  return {
+    evaluateReadAccess: vi.fn().mockResolvedValue(result),
+  };
+}
+
+describe('FederationScopeService', () => {
+  let service: FederationScopeService;
+
+  beforeEach(() => {
+    service = new FederationScopeService();
+  });
+
+  it('allows a granted resource and returns a capped query filter', async () => {
+    const nativeRbac = makeNativeRbac({
+      allowed: true,
+      access: { includePersonal: true, teamIds: ['team-1', 'team-2'] },
+    });
+
+    const result = await service.evaluateAccess({
+      context: makeContext({
+        resources: ['tasks'],
+        filters: { tasks: { include_teams: ['team-1', 'team-3'], include_personal: true } },
+        max_rows_per_query: 50,
+      }),
+      resource: 'tasks',
+      requestedLimit: 500,
+      nativeRbac,
+    });
+
+    expect(result).toEqual({
+      allowed: true,
+      filter: {
+        resource: 'tasks',
+        subjectUserId: SUBJECT_USER_ID,
+        includePersonal: true,
+        teamIds: ['team-1'],
+        limit: 50,
+        maxRowsPerQuery: 50,
+      },
+    });
+    expect(nativeRbac.evaluateReadAccess).toHaveBeenCalledWith({
+      grantId: GRANT_ID,
+      peerId: PEER_ID,
+      subjectUserId: SUBJECT_USER_ID,
+      resource: 'tasks',
+    });
+  });
+
+  it('defaults absent resource filters to native RBAC personal and team visibility', async () => {
+    const result = await service.evaluateAccess({
+      context: makeContext({ resources: ['notes'], max_rows_per_query: 100 }),
+      resource: 'notes',
+      nativeRbac: makeNativeRbac({
+        allowed: true,
+        access: { includePersonal: true, teamIds: ['team-1', 'team-2'] },
+      }),
+    });
+
+    expect(result).toMatchObject({
+      allowed: true,
+      filter: {
+        includePersonal: true,
+        teamIds: ['team-1', 'team-2'],
+        limit: 100,
+      },
+    });
+  });
+
+  it('honors include_personal false even when native RBAC allows personal rows', async () => {
+    const result = await service.evaluateAccess({
+      context: makeContext({
+        resources: ['memory'],
+        filters: { memory: { include_personal: false } },
+        max_rows_per_query: 25,
+      }),
+      resource: 'memory',
+      nativeRbac: makeNativeRbac({
+        allowed: true,
+        access: { includePersonal: true, teamIds: [] },
+      }),
+    });
+
+    expect(result).toMatchObject({
+      allowed: true,
+      filter: {
+        includePersonal: false,
+        teamIds: [],
+      },
+    });
+  });
+
+  it('does not leak personal rows when scope allows personal but native RBAC denies personal', async () => {
+    const result = await service.evaluateAccess({
+      context: makeContext({
+        resources: ['tasks'],
+        filters: { tasks: { include_personal: true } },
+        max_rows_per_query: 25,
+      }),
+      resource: 'tasks',
+      nativeRbac: makeNativeRbac({
+        allowed: true,
+        access: { includePersonal: false, teamIds: ['team-1'] },
+      }),
+    });
+
+    expect(result).toMatchObject({
+      allowed: true,
+      filter: {
+        includePersonal: false,
+        teamIds: ['team-1'],
+      },
+    });
+  });
+
+  it('does not widen native RBAC when scope includes teams the user cannot access', async () => {
+    const result = await service.evaluateAccess({
+      context: makeContext({
+        resources: ['tasks'],
+        filters: { tasks: { include_teams: ['team-2'], include_personal: false } },
+        max_rows_per_query: 25,
+      }),
+      resource: 'tasks',
+      nativeRbac: makeNativeRbac({
+        allowed: true,
+        access: { includePersonal: true, teamIds: ['team-1'] },
+      }),
+    });
+
+    expect(result).toMatchObject({
+      allowed: true,
+      filter: {
+        includePersonal: false,
+        teamIds: [],
+      },
+    });
+  });
+
+  it('denies invalid grant scope before RBAC evaluation', async () => {
+    const nativeRbac = makeNativeRbac({
+      allowed: true,
+      access: { includePersonal: true, teamIds: [] },
+    });
+
+    const result = await service.evaluateAccess({
+      context: makeContext({ resources: [], max_rows_per_query: 100 }),
+      resource: 'tasks',
+      nativeRbac,
+    });
+
+    expect(result).toMatchObject({
+      allowed: false,
+      deny: {
+        code: 'invalid_scope',
+        stage: 'scope_parse',
+        statusCode: 400,
+        grantId: GRANT_ID,
+        subjectUserId: SUBJECT_USER_ID,
+        resource: 'tasks',
+      },
+    });
+    expect(nativeRbac.evaluateReadAccess).not.toHaveBeenCalled();
+  });
+
+  it('denies unsupported resource names before RBAC evaluation', async () => {
+    const nativeRbac = makeNativeRbac({
+      allowed: true,
+      access: { includePersonal: true, teamIds: [] },
+    });
+
+    const result = await service.evaluateAccess({
+      context: makeContext({ resources: ['tasks'], max_rows_per_query: 100 }),
+      resource: 'unknown_resource',
+      nativeRbac,
+    });
+
+    expect(result).toMatchObject({
+      allowed: false,
+      deny: {
+        code: 'invalid_resource',
+        stage: 'resource_allowlist',
+        statusCode: 403,
+      },
+    });
+    expect(nativeRbac.evaluateReadAccess).not.toHaveBeenCalled();
+  });
+
+  it('denies resources explicitly present in excluded_resources before allowlist miss', async () => {
+    const nativeRbac = makeNativeRbac({
+      allowed: true,
+      access: { includePersonal: true, teamIds: [] },
+    });
+
+    const result = await service.evaluateAccess({
+      context: makeContext({
+        resources: ['tasks'],
+        excluded_resources: ['credentials'],
+        max_rows_per_query: 100,
+      }),
+      resource: 'credentials',
+      nativeRbac,
+    });
+
+    expect(result).toMatchObject({
+      allowed: false,
+      deny: {
+        code: 'resource_excluded',
+        stage: 'resource_exclusion',
+        statusCode: 403,
+        resource: 'credentials',
+      },
+    });
+    expect(nativeRbac.evaluateReadAccess).not.toHaveBeenCalled();
+  });
+
+  it('denies supported resources that are not granted by scope', async () => {
+    const nativeRbac = makeNativeRbac({
+      allowed: true,
+      access: { includePersonal: true, teamIds: [] },
+    });
+
+    const result = await service.evaluateAccess({
+      context: makeContext({ resources: ['tasks'], max_rows_per_query: 100 }),
+      resource: 'notes',
+      nativeRbac,
+    });
+
+    expect(result).toMatchObject({
+      allowed: false,
+      deny: {
+        code: 'resource_not_granted',
+        stage: 'resource_allowlist',
+        statusCode: 403,
+        resource: 'notes',
+      },
+    });
+    expect(nativeRbac.evaluateReadAccess).not.toHaveBeenCalled();
+  });
+
+  it('denies invalid requested row limits before RBAC evaluation', async () => {
+    const nativeRbac = makeNativeRbac({
+      allowed: true,
+      access: { includePersonal: true, teamIds: [] },
+    });
+
+    const result = await service.evaluateAccess({
+      context: makeContext({ resources: ['tasks'], max_rows_per_query: 100 }),
+      resource: 'tasks',
+      requestedLimit: 0,
+      nativeRbac,
+    });
+
+    expect(result).toMatchObject({
+      allowed: false,
+      deny: {
+        code: 'invalid_limit',
+        stage: 'row_cap',
+        statusCode: 400,
+        details: { requestedLimit: 0 },
+      },
+    });
+    expect(nativeRbac.evaluateReadAccess).not.toHaveBeenCalled();
+  });
+
+  it('denies when native RBAC rejects subjectUserId access to the resource', async () => {
+    const result = await service.evaluateAccess({
+      context: makeContext({ resources: ['tasks'], max_rows_per_query: 100 }),
+      resource: 'tasks',
+      nativeRbac: makeNativeRbac({
+        allowed: false,
+        reason: 'read:tasks denied',
+        details: { permission: 'tasks:read' },
+      }),
+    });
+
+    expect(result).toEqual({
+      allowed: false,
+      deny: {
+        code: 'native_rbac_denied',
+        stage: 'native_rbac',
+        statusCode: 403,
+        message: 'read:tasks denied',
+        grantId: GRANT_ID,
+        peerId: PEER_ID,
+        subjectUserId: SUBJECT_USER_ID,
+        resource: 'tasks',
+        details: { permission: 'tasks:read' },
+      },
+    });
+  });
+});

apps/gateway/src/federation/server/index.ts

@@ -10,4 +10,22 @@
  */
 
 export { FederationAuthGuard } from './federation-auth.guard.js';
+export { FederationScopeService } from './scope.service.js';
 export type { FederationContext } from './federation-context.js';
+export type {
+  FederationNativeRbacAccess,
+  FederationNativeRbacAllowedResult,
+  FederationNativeRbacDeniedResult,
+  FederationNativeRbacEvaluator,
+  FederationNativeRbacRequest,
+  FederationNativeRbacResult,
+  FederationScopeAllowedResult,
+  FederationScopeDeniedResult,
+  FederationScopeDenyCode,
+  FederationScopeDenyDetails,
+  FederationScopeDenyReason,
+  FederationScopeDenyStage,
+  FederationScopeEvaluationInput,
+  FederationScopeEvaluationResult,
+  FederationScopeQueryFilter,
+} from './scope.service.js';

apps/gateway/src/federation/server/scope.service.ts

@@ -0,0 +1,272 @@
+/**
+ * FederationScopeService — M3 server-side scope enforcement pipeline.
+ *
+ * Pure trust-boundary service: it validates the grant scope, asks an injected
+ * native RBAC evaluator what the subject user can read locally, intersects that
+ * answer with the federation scope filters, and returns a query filter for the
+ * verb controllers. The service performs no DB calls directly.
+ */
+
+import { Injectable } from '@nestjs/common';
+import {
+  FEDERATION_RESOURCE_VALUES,
+  type FederationResource,
+  FederationScopeError,
+  parseFederationScope,
+} from '../scope-schema.js';
+import type { FederationContext } from './federation-context.js';
+
+const federationResourceSet: ReadonlySet<string> = new Set<string>(FEDERATION_RESOURCE_VALUES);
+
+export type FederationScopeDenyStage =
+  | 'scope_parse'
+  | 'resource_allowlist'
+  | 'resource_exclusion'
+  | 'native_rbac'
+  | 'row_cap';
+
+export type FederationScopeDenyCode =
+  | 'invalid_scope'
+  | 'invalid_resource'
+  | 'resource_not_granted'
+  | 'resource_excluded'
+  | 'native_rbac_denied'
+  | 'invalid_limit';
+
+export type FederationScopeDenyStatus = 400 | 403;
+
+export interface FederationScopeDenyDetails {
+  readonly [key: string]: string | number | boolean | readonly string[];
+}
+
+export interface FederationScopeDenyReason {
+  readonly code: FederationScopeDenyCode;
+  readonly stage: FederationScopeDenyStage;
+  readonly statusCode: FederationScopeDenyStatus;
+  readonly message: string;
+  readonly grantId: string;
+  readonly peerId: string;
+  readonly subjectUserId: string;
+  readonly resource: string;
+  readonly details?: FederationScopeDenyDetails;
+}
+
+export interface FederationNativeRbacRequest {
+  readonly grantId: string;
+  readonly peerId: string;
+  readonly subjectUserId: string;
+  readonly resource: FederationResource;
+}
+
+export interface FederationNativeRbacAccess {
+  /** Whether this user may read personal rows for this resource. */
+  readonly includePersonal: boolean;
+
+  /** Team IDs this user may read for this resource under native RBAC. */
+  readonly teamIds: readonly string[];
+}
+
+export interface FederationNativeRbacAllowedResult {
+  readonly allowed: true;
+  readonly access: FederationNativeRbacAccess;
+}
+
+export interface FederationNativeRbacDeniedResult {
+  readonly allowed: false;
+  readonly reason?: string;
+  readonly details?: FederationScopeDenyDetails;
+}
+
+export type FederationNativeRbacResult =
+  | FederationNativeRbacAllowedResult
+  | FederationNativeRbacDeniedResult;
+
+export interface FederationNativeRbacEvaluator {
+  evaluateReadAccess(request: FederationNativeRbacRequest): Promise<FederationNativeRbacResult>;
+}
+
+export interface FederationScopeEvaluationInput {
+  readonly context: FederationContext;
+  readonly resource: string;
+  readonly requestedLimit?: number;
+  readonly nativeRbac: FederationNativeRbacEvaluator;
+}
+
+export interface FederationScopeQueryFilter {
+  readonly resource: FederationResource;
+  readonly subjectUserId: string;
+  readonly includePersonal: boolean;
+  readonly teamIds: readonly string[];
+  readonly limit: number;
+  readonly maxRowsPerQuery: number;
+}
+
+export interface FederationScopeAllowedResult {
+  readonly allowed: true;
+  readonly filter: FederationScopeQueryFilter;
+}
+
+export interface FederationScopeDeniedResult {
+  readonly allowed: false;
+  readonly deny: FederationScopeDenyReason;
+}
+
+export type FederationScopeEvaluationResult =
+  | FederationScopeAllowedResult
+  | FederationScopeDeniedResult;
+
+function isFederationResource(resource: string): resource is FederationResource {
+  return federationResourceSet.has(resource);
+}
+
+function uniqueStrings(values: readonly string[]): readonly string[] {
+  return Array.from(new Set<string>(values));
+}
+
+function intersectTeamIds(
+  nativeTeamIds: readonly string[],
+  scopedTeamIds: readonly string[] | undefined,
+): readonly string[] {
+  const uniqueNativeTeamIds = uniqueStrings(nativeTeamIds);
+
+  if (scopedTeamIds === undefined) {
+    return uniqueNativeTeamIds;
+  }
+
+  const nativeSet = new Set<string>(uniqueNativeTeamIds);
+  return uniqueStrings(scopedTeamIds).filter((teamId: string): boolean => nativeSet.has(teamId));
+}
+
+function makeDenyReason(params: {
+  readonly code: FederationScopeDenyCode;
+  readonly stage: FederationScopeDenyStage;
+  readonly statusCode?: FederationScopeDenyStatus;
+  readonly message: string;
+  readonly context: FederationContext;
+  readonly resource: string;
+  readonly details?: FederationScopeDenyDetails;
+}): FederationScopeDeniedResult {
+  return {
+    allowed: false,
+    deny: {
+      code: params.code,
+      stage: params.stage,
+      statusCode: params.statusCode ?? 403,
+      message: params.message,
+      grantId: params.context.grantId,
+      peerId: params.context.peerId,
+      subjectUserId: params.context.subjectUserId,
+      resource: params.resource,
+      ...(params.details !== undefined ? { details: params.details } : {}),
+    },
+  };
+}
+
+@Injectable()
+export class FederationScopeService {
+  async evaluateAccess(
+    input: FederationScopeEvaluationInput,
+  ): Promise<FederationScopeEvaluationResult> {
+    const { context, resource, requestedLimit, nativeRbac } = input;
+
+    let scope: ReturnType<typeof parseFederationScope>;
+    try {
+      scope = parseFederationScope(context.scope);
+    } catch (error: unknown) {
+      const message =
+        error instanceof FederationScopeError
+          ? 'Federation grant scope is invalid'
+          : 'Federation grant scope could not be parsed';
+      const details = error instanceof Error ? { reason: error.message } : undefined;
+      return makeDenyReason({
+        code: 'invalid_scope',
+        stage: 'scope_parse',
+        statusCode: 400,
+        message,
+        context,
+        resource,
+        ...(details !== undefined ? { details } : {}),
+      });
+    }
+
+    if (!isFederationResource(resource)) {
+      return makeDenyReason({
+        code: 'invalid_resource',
+        stage: 'resource_allowlist',
+        message: 'Requested federation resource is not supported',
+        context,
+        resource,
+        details: { supportedResources: FEDERATION_RESOURCE_VALUES },
+      });
+    }
+
+    if (scope.excluded_resources.includes(resource)) {
+      return makeDenyReason({
+        code: 'resource_excluded',
+        stage: 'resource_exclusion',
+        message: 'Requested federation resource is explicitly excluded by grant scope',
+        context,
+        resource,
+      });
+    }
+
+    if (!scope.resources.includes(resource)) {
+      return makeDenyReason({
+        code: 'resource_not_granted',
+        stage: 'resource_allowlist',
+        message: 'Requested federation resource is not granted by scope',
+        context,
+        resource,
+        details: { grantedResources: scope.resources },
+      });
+    }
+
+    if (requestedLimit !== undefined && (!Number.isInteger(requestedLimit) || requestedLimit < 1)) {
+      return makeDenyReason({
+        code: 'invalid_limit',
+        stage: 'row_cap',
+        statusCode: 400,
+        message: 'Requested row limit must be a positive integer',
+        context,
+        resource,
+        details: { requestedLimit },
+      });
+    }
+
+    const nativeResult = await nativeRbac.evaluateReadAccess({
+      grantId: context.grantId,
+      peerId: context.peerId,
+      subjectUserId: context.subjectUserId,
+      resource,
+    });
+
+    if (!nativeResult.allowed) {
+      return makeDenyReason({
+        code: 'native_rbac_denied',
+        stage: 'native_rbac',
+        message: nativeResult.reason ?? 'Subject user is not allowed to read this resource',
+        context,
+        resource,
+        ...(nativeResult.details !== undefined ? { details: nativeResult.details } : {}),
+      });
+    }
+
+    const scopeFilter = scope.filters?.[resource];
+    const includePersonal =
+      Boolean(scopeFilter?.include_personal ?? true) && nativeResult.access.includePersonal;
+    const teamIds = intersectTeamIds(nativeResult.access.teamIds, scopeFilter?.include_teams);
+    const limit = Math.min(requestedLimit ?? scope.max_rows_per_query, scope.max_rows_per_query);
+
+    return {
+      allowed: true,
+      filter: {
+        resource,
+        subjectUserId: context.subjectUserId,
+        includePersonal,
+        teamIds,
+        limit,
+        maxRowsPerQuery: scope.max_rows_per_query,
+      },
+    };
+  }
+}

apps/gateway/src/federation/server/verbs/__tests__/capabilities.controller.spec.ts

@@ -0,0 +1,88 @@
+import 'reflect-metadata';
+import { RequestMethod } from '@nestjs/common';
+import { describe, expect, it } from 'vitest';
+import type { FastifyRequest } from 'fastify';
+import { FederationCapabilitiesResponseSchema, FEDERATION_VERBS } from '@mosaicstack/types';
+import { FederationScopeError } from '../../../scope-schema.js';
+import { FederationAuthGuard } from '../../federation-auth.guard.js';
+import { CapabilitiesController } from '../capabilities.controller.js';
+
+const VALID_SCOPE = {
+  resources: ['tasks', 'notes'],
+  excluded_resources: ['credentials'],
+  max_rows_per_query: 250,
+} as const;
+
+const DEFAULTED_SCOPE = {
+  resources: ['memory'],
+  max_rows_per_query: 10,
+} as const;
+
+function makeRequest(scope: Record<string, unknown>): FastifyRequest {
+  return {
+    federationContext: {
+      grantId: 'grant-1',
+      peerId: 'peer-1',
+      subjectUserId: 'user-1',
+      scope,
+    },
+  } as FastifyRequest;
+}
+
+describe('CapabilitiesController', () => {
+  it('declares GET /api/federation/v1/capabilities', () => {
+    expect(Reflect.getMetadata('path', CapabilitiesController)).toBe(
+      'api/federation/v1/capabilities',
+    );
+    expect(Reflect.getMetadata('path', CapabilitiesController.prototype.getCapabilities)).toBe('/');
+    expect(Reflect.getMetadata('method', CapabilitiesController.prototype.getCapabilities)).toBe(
+      RequestMethod.GET,
+    );
+  });
+
+  it('is protected only by FederationAuthGuard', () => {
+    const guards = Reflect.getMetadata('__guards__', CapabilitiesController) as unknown[];
+
+    expect(guards).toEqual([FederationAuthGuard]);
+  });
+
+  it('returns resources, excluded resources, max rows, and M3 supported verbs from the active grant scope', () => {
+    const controller = new CapabilitiesController();
+
+    const response = controller.getCapabilities(makeRequest(VALID_SCOPE));
+
+    expect(response).toEqual({
+      resources: ['tasks', 'notes'],
+      excluded_resources: ['credentials'],
+      max_rows_per_query: 250,
+      supported_verbs: [...FEDERATION_VERBS],
+    });
+    expect(FederationCapabilitiesResponseSchema.safeParse(response).success).toBe(true);
+  });
+
+  it('applies scope defaults without RBAC or resource filtering', () => {
+    const controller = new CapabilitiesController();
+
+    const response = controller.getCapabilities(makeRequest(DEFAULTED_SCOPE));
+
+    expect(response).toEqual({
+      resources: ['memory'],
+      excluded_resources: [],
+      max_rows_per_query: 10,
+      supported_verbs: ['list', 'get', 'capabilities'],
+    });
+  });
+
+  it('rejects invalid scope state instead of returning an invalid capabilities contract', () => {
+    const controller = new CapabilitiesController();
+
+    expect(() =>
+      controller.getCapabilities(
+        makeRequest({
+          resources: [],
+          max_rows_per_query: 0,
+        }),
+      ),
+    ).toThrow(FederationScopeError);
+  });
+});

apps/gateway/src/federation/server/verbs/__tests__/list-query.service.spec.ts

@@ -0,0 +1,428 @@
+import { afterAll, beforeAll, describe, expect, it, vi } from 'vitest';
+import {
+  createPgliteDb,
+  insights,
+  missionTasks,
+  missions,
+  preferences,
+  projects,
+  runPgliteMigrations,
+  teams,
+  users,
+  type Db,
+  type DbHandle,
+} from '@mosaicstack/db';
+import type { FederationScopeQueryFilter } from '../../scope.service.js';
+import { FederationListQueryService } from '../list-query.service.js';
+
+const TASK_FILTER: FederationScopeQueryFilter = {
+  resource: 'tasks',
+  subjectUserId: 'user-1',
+  includePersonal: true,
+  teamIds: [],
+  limit: 2,
+  maxRowsPerQuery: 2,
+};
+
+const SUBJECT_USER_ID = 'fed-m3-05-subject';
+const OTHER_USER_ID = 'fed-m3-05-other';
+const TEAM_ID = '05000000-0000-4000-8000-000000000001';
+const UNAUTHORIZED_TEAM_ID = '05000000-0000-4000-8000-000000000002';
+const PERSONAL_PROJECT_ID = '05000000-0000-4000-8000-000000000101';
+const TEAM_PROJECT_ID = '05000000-0000-4000-8000-000000000102';
+const UNAUTHORIZED_PROJECT_ID = '05000000-0000-4000-8000-000000000103';
+const PERSONAL_MISSION_ID = '05000000-0000-4000-8000-000000000201';
+const TEAM_MISSION_ID = '05000000-0000-4000-8000-000000000202';
+const UNAUTHORIZED_MISSION_ID = '05000000-0000-4000-8000-000000000203';
+const SUBJECT_TEAM_NOTE_ID = '05000000-0000-4000-8000-000000000301';
+const OTHER_TEAM_NOTE_ID = '05000000-0000-4000-8000-000000000302';
+const SUBJECT_PERSONAL_NOTE_ID = '05000000-0000-4000-8000-000000000303';
+const SUBJECT_UNAUTHORIZED_NOTE_ID = '05000000-0000-4000-8000-000000000304';
+const INSIGHT_ONE_ID = '05000000-0000-4000-8000-000000000401';
+const INSIGHT_TWO_ID = '05000000-0000-4000-8000-000000000402';
+const PREFERENCE_ONE_ID = '05000000-0000-4000-8000-000000000501';
+const PREFERENCE_TWO_ID = '05000000-0000-4000-8000-000000000502';
+
+let dbHandle: DbHandle | undefined;
+
+function makeService() {
+  return new FederationListQueryService({} as Db);
+}
+
+function makeDbService() {
+  if (!dbHandle) {
+    throw new Error('test DB not initialized');
+  }
+  return new FederationListQueryService(dbHandle.db);
+}
+
+async function seedNotesFixture() {
+  if (!dbHandle) {
+    throw new Error('test DB not initialized');
+  }
+
+  await dbHandle.db.insert(users).values([
+    {
+      id: SUBJECT_USER_ID,
+      name: 'Federation Subject',
+      email: `${SUBJECT_USER_ID}@example.test`,
+      emailVerified: false,
+    },
+    {
+      id: OTHER_USER_ID,
+      name: 'Federation Other',
+      email: `${OTHER_USER_ID}@example.test`,
+      emailVerified: false,
+    },
+  ]);
+
+  await dbHandle.db.insert(teams).values([
+    {
+      id: TEAM_ID,
+      name: 'FED-M3-05 Team',
+      slug: 'fed-m3-05-team',
+      ownerId: SUBJECT_USER_ID,
+      managerId: SUBJECT_USER_ID,
+    },
+    {
+      id: UNAUTHORIZED_TEAM_ID,
+      name: 'FED-M3-05 Unauthorized Team',
+      slug: 'fed-m3-05-unauthorized-team',
+      ownerId: OTHER_USER_ID,
+      managerId: OTHER_USER_ID,
+    },
+  ]);
+
+  await dbHandle.db.insert(projects).values([
+    {
+      id: PERSONAL_PROJECT_ID,
+      name: 'FED-M3-05 Personal Project',
+      ownerId: SUBJECT_USER_ID,
+      ownerType: 'user',
+    },
+    {
+      id: TEAM_PROJECT_ID,
+      name: 'FED-M3-05 Team Project',
+      teamId: TEAM_ID,
+      ownerType: 'team',
+    },
+    {
+      id: UNAUTHORIZED_PROJECT_ID,
+      name: 'FED-M3-05 Unauthorized Project',
+      teamId: UNAUTHORIZED_TEAM_ID,
+      ownerType: 'team',
+    },
+  ]);
+
+  await dbHandle.db.insert(missions).values([
+    {
+      id: PERSONAL_MISSION_ID,
+      name: 'FED-M3-05 Personal Mission',
+      projectId: PERSONAL_PROJECT_ID,
+      userId: SUBJECT_USER_ID,
+    },
+    {
+      id: TEAM_MISSION_ID,
+      name: 'FED-M3-05 Team Mission',
+      projectId: TEAM_PROJECT_ID,
+      userId: SUBJECT_USER_ID,
+    },
+    {
+      id: UNAUTHORIZED_MISSION_ID,
+      name: 'FED-M3-05 Unauthorized Mission',
+      projectId: UNAUTHORIZED_PROJECT_ID,
+      userId: SUBJECT_USER_ID,
+    },
+  ]);
+
+  await dbHandle.db.insert(missionTasks).values([
+    {
+      id: SUBJECT_TEAM_NOTE_ID,
+      missionId: TEAM_MISSION_ID,
+      userId: SUBJECT_USER_ID,
+      notes: 'subject note on team mission',
+      createdAt: new Date('2026-06-24T03:00:00.000Z'),
+      updatedAt: new Date('2026-06-24T03:00:00.000Z'),
+    },
+    {
+      id: OTHER_TEAM_NOTE_ID,
+      missionId: TEAM_MISSION_ID,
+      userId: OTHER_USER_ID,
+      notes: 'other user note on team mission',
+      createdAt: new Date('2026-06-24T02:00:00.000Z'),
+      updatedAt: new Date('2026-06-24T02:00:00.000Z'),
+    },
+    {
+      id: SUBJECT_PERSONAL_NOTE_ID,
+      missionId: PERSONAL_MISSION_ID,
+      userId: SUBJECT_USER_ID,
+      notes: 'subject note on personal mission',
+      createdAt: new Date('2026-06-24T01:00:00.000Z'),
+      updatedAt: new Date('2026-06-24T01:00:00.000Z'),
+    },
+    {
+      id: SUBJECT_UNAUTHORIZED_NOTE_ID,
+      missionId: UNAUTHORIZED_MISSION_ID,
+      userId: SUBJECT_USER_ID,
+      notes: 'subject note outside grant-visible missions',
+      createdAt: new Date('2026-06-24T04:00:00.000Z'),
+      updatedAt: new Date('2026-06-24T04:00:00.000Z'),
+    },
+  ]);
+
+  const memoryCreatedAt = new Date('2026-06-24T05:00:00.000Z');
+  await dbHandle.db.insert(insights).values([
+    {
+      id: INSIGHT_ONE_ID,
+      userId: SUBJECT_USER_ID,
+      content: 'first insight',
+      source: 'agent',
+      createdAt: memoryCreatedAt,
+      updatedAt: memoryCreatedAt,
+    },
+    {
+      id: INSIGHT_TWO_ID,
+      userId: SUBJECT_USER_ID,
+      content: 'second insight',
+      source: 'agent',
+      createdAt: memoryCreatedAt,
+      updatedAt: memoryCreatedAt,
+    },
+  ]);
+
+  await dbHandle.db.insert(preferences).values([
+    {
+      id: PREFERENCE_ONE_ID,
+      userId: SUBJECT_USER_ID,
+      key: 'fed-m3-05-pref-1',
+      value: { enabled: true },
+      createdAt: memoryCreatedAt,
+      updatedAt: memoryCreatedAt,
+    },
+    {
+      id: PREFERENCE_TWO_ID,
+      userId: SUBJECT_USER_ID,
+      key: 'fed-m3-05-pref-2',
+      value: { enabled: false },
+      createdAt: memoryCreatedAt,
+      updatedAt: memoryCreatedAt,
+    },
+  ]);
+}
+
+function stubRows(
+  service: FederationListQueryService,
+  ...pages: Array<Array<Record<string, unknown>>>
+) {
+  const mock = vi.fn();
+  for (const page of pages) {
+    mock.mockResolvedValueOnce(page);
+  }
+  (
+    service as unknown as {
+      listAllRows: (
+        _filter: FederationScopeQueryFilter,
+        _rowLimit: number,
+        _cursor: unknown,
+      ) => Promise<Array<Record<string, unknown>>>;
+    }
+  ).listAllRows = mock;
+  return mock;
+}
+
+describe('FederationListQueryService', () => {
+  beforeAll(async () => {
+    dbHandle = createPgliteDb(`memory://fed-m3-05-list-${Date.now()}`);
+    await runPgliteMigrations(dbHandle);
+    await seedNotesFixture();
+  });
+
+  afterAll(async () => {
+    await dbHandle?.close();
+    dbHandle = undefined;
+  });
+
+  it('denies sensitive resources in native RBAC for M3 list reads', async () => {
+    const service = makeService();
+
+    await expect(
+      service.evaluateReadAccess({
+        grantId: 'grant-1',
+        peerId: 'peer-1',
+        subjectUserId: 'user-1',
+        resource: 'credentials',
+      }),
+    ).resolves.toMatchObject({
+      allowed: false,
+      reason: 'credentials federation list access is not implemented in M3',
+    });
+  });
+
+  it('allows personal memory reads without requiring team lookup', async () => {
+    const service = makeService();
+
+    await expect(
+      service.evaluateReadAccess({
+        grantId: 'grant-1',
+        peerId: 'peer-1',
+        subjectUserId: 'user-1',
+        resource: 'memory',
+      }),
+    ).resolves.toEqual({
+      allowed: true,
+      access: { includePersonal: true, teamIds: [] },
+    });
+  });
+
+  it('applies the scope row cap and returns an opaque next cursor when truncated', async () => {
+    const service = makeService();
+    const listAllRows = stubRows(
+      service,
+      [
+        { id: '3', createdAt: new Date('2026-06-24T03:00:00.000Z') },
+        { id: '2', createdAt: new Date('2026-06-24T02:00:00.000Z') },
+        { id: '1', createdAt: new Date('2026-06-24T01:00:00.000Z') },
+      ],
+      [{ id: '1', createdAt: new Date('2026-06-24T01:00:00.000Z') }],
+    );
+
+    const firstPage = await service.list({ filter: TASK_FILTER });
+
+    expect(firstPage).toEqual({
+      items: [
+        { id: '3', createdAt: new Date('2026-06-24T03:00:00.000Z') },
+        { id: '2', createdAt: new Date('2026-06-24T02:00:00.000Z') },
+      ],
+      truncated: true,
+      nextCursor: expect.any(String),
+    });
+
+    expect(listAllRows).toHaveBeenNthCalledWith(1, TASK_FILTER, 3, undefined);
+
+    const secondPage = await service.list({ filter: TASK_FILTER, cursor: firstPage.nextCursor });
+    expect(secondPage).toEqual({
+      items: [{ id: '1', createdAt: new Date('2026-06-24T01:00:00.000Z') }],
+      truncated: false,
+    });
+    expect(listAllRows).toHaveBeenNthCalledWith(
+      2,
+      TASK_FILTER,
+      3,
+      expect.objectContaining({ id: '2' }),
+    );
+  });
+
+  it('rejects invalid cursors instead of falling back to the first page', async () => {
+    const service = makeService();
+    stubRows(service, [{ id: '1' }]);
+
+    await expect(service.list({ filter: TASK_FILTER, cursor: 'not-base64-json' })).rejects.toThrow(
+      'Invalid federation list cursor',
+    );
+  });
+
+  it('throws when a truncated page cannot encode a resumable cursor', async () => {
+    const service = makeService();
+    stubRows(service, [
+      { id: '2', createdAt: 'not-a-date' },
+      { id: '1', createdAt: 'not-a-date' },
+    ]);
+
+    await expect(service.list({ filter: { ...TASK_FILTER, limit: 1 } })).rejects.toThrow(
+      'Federation list cursor cannot be encoded',
+    );
+  });
+
+  it('throws on unsupported resources instead of crashing pagination', async () => {
+    const service = makeService();
+
+    await expect(
+      service.list({
+        filter: {
+          ...TASK_FILTER,
+          resource: 'unknown-resource' as FederationScopeQueryFilter['resource'],
+        },
+      }),
+    ).rejects.toThrow('Unsupported federation list resource');
+  });
+
+  it('does not leak another user mission task notes through team-scoped note reads', async () => {
+    const service = makeDbService();
+
+    const result = await service.list({
+      filter: {
+        resource: 'notes',
+        subjectUserId: SUBJECT_USER_ID,
+        includePersonal: false,
+        teamIds: [TEAM_ID],
+        limit: 10,
+        maxRowsPerQuery: 10,
+      },
+    });
+
+    const ids = result.items.map((item) => item['id']);
+    expect(ids).toEqual([SUBJECT_TEAM_NOTE_ID]);
+    expect(ids).not.toContain(OTHER_TEAM_NOTE_ID);
+  });
+
+  it('does not return subject personal mission task notes when includePersonal is false', async () => {
+    const service = makeDbService();
+
+    const result = await service.list({
+      filter: {
+        resource: 'notes',
+        subjectUserId: SUBJECT_USER_ID,
+        includePersonal: false,
+        teamIds: [TEAM_ID],
+        limit: 10,
+        maxRowsPerQuery: 10,
+      },
+    });
+
+    expect(result.items.map((item) => item['id'])).not.toContain(SUBJECT_PERSONAL_NOTE_ID);
+  });
+
+  it('does not return subject notes from missions outside the grant-visible project set', async () => {
+    const service = makeDbService();
+
+    const result = await service.list({
+      filter: {
+        resource: 'notes',
+        subjectUserId: SUBJECT_USER_ID,
+        includePersonal: true,
+        teamIds: [TEAM_ID],
+        limit: 10,
+        maxRowsPerQuery: 10,
+      },
+    });
+
+    const ids = result.items.map((item) => item['id']);
+    expect(ids).toContain(SUBJECT_PERSONAL_NOTE_ID);
+    expect(ids).toContain(SUBJECT_TEAM_NOTE_ID);
+    expect(ids).not.toContain(SUBJECT_UNAUTHORIZED_NOTE_ID);
+    expect(ids).not.toContain(OTHER_TEAM_NOTE_ID);
+  });
+
+  it('paginates memory deterministically across insights and preferences', async () => {
+    const service = makeDbService();
+    const filter: FederationScopeQueryFilter = {
+      resource: 'memory',
+      subjectUserId: SUBJECT_USER_ID,
+      includePersonal: true,
+      teamIds: [],
+      limit: 2,
+      maxRowsPerQuery: 2,
+    };
+
+    const firstPage = await service.list({ filter });
+    const secondPage = await service.list({ filter, cursor: firstPage.nextCursor });
+    const firstPageIds = firstPage.items.map((item) => item['id']);
+    const secondPageIds = secondPage.items.map((item) => item['id']);
+    const allIds = [...firstPageIds, ...secondPageIds];
+
+    expect(firstPage).toMatchObject({ truncated: true, nextCursor: expect.any(String) });
+    expect(firstPageIds).toEqual([INSIGHT_TWO_ID, INSIGHT_ONE_ID]);
+    expect(secondPageIds).toEqual([PREFERENCE_TWO_ID, PREFERENCE_ONE_ID]);
+    expect(new Set(allIds).size).toBe(allIds.length);
+  });
+});

apps/gateway/src/federation/server/verbs/__tests__/list.controller.spec.ts

@@ -0,0 +1,188 @@
+import 'reflect-metadata';
+import { RequestMethod } from '@nestjs/common';
+import type { FastifyRequest } from 'fastify';
+import { beforeEach, describe, expect, it, vi } from 'vitest';
+import { FederationAuthGuard } from '../../federation-auth.guard.js';
+import type {
+  FederationScopeEvaluationResult,
+  FederationScopeQueryFilter,
+} from '../../scope.service.js';
+import { ListController } from '../list.controller.js';
+import type { FederationListQueryResult } from '../list-query.service.js';
+
+const FEDERATION_CONTEXT = {
+  grantId: 'grant-1',
+  peerId: 'peer-1',
+  subjectUserId: 'user-1',
+  scope: { resources: ['tasks'], max_rows_per_query: 25 },
+};
+
+const TASK_FILTER: FederationScopeQueryFilter = {
+  resource: 'tasks',
+  subjectUserId: 'user-1',
+  includePersonal: true,
+  teamIds: ['team-1'],
+  limit: 10,
+  maxRowsPerQuery: 25,
+};
+
+function makeRequest(): FastifyRequest {
+  return { federationContext: FEDERATION_CONTEXT } as unknown as FastifyRequest;
+}
+
+function allowedScope(
+  filter: FederationScopeQueryFilter = TASK_FILTER,
+): FederationScopeEvaluationResult {
+  return { allowed: true, filter };
+}
+
+function makeController(opts?: {
+  scopeResult?: FederationScopeEvaluationResult;
+  queryResult?: FederationListQueryResult;
+}) {
+  const scope = {
+    evaluateAccess: vi.fn().mockResolvedValue(opts?.scopeResult ?? allowedScope()),
+  };
+  const query = {
+    evaluateReadAccess: vi.fn(),
+    list: vi.fn().mockResolvedValue(
+      opts?.queryResult ?? {
+        items: [
+          {
+            id: 'task-1',
+            title: 'Federated task',
+            createdAt: new Date('2026-06-24T00:00:00.000Z'),
+          },
+        ],
+        truncated: false,
+      },
+    ),
+  };
+
+  return {
+    controller: new ListController(scope as never, query as never),
+    scope,
+    query,
+  };
+}
+
+describe('ListController', () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+  });
+
+  it('declares POST /api/federation/v1/list/:resource protected only by FederationAuthGuard', () => {
+    expect(Reflect.getMetadata('path', ListController)).toBe('api/federation/v1/list');
+    expect(Reflect.getMetadata('path', ListController.prototype.list)).toBe(':resource');
+    expect(Reflect.getMetadata('method', ListController.prototype.list)).toBe(RequestMethod.POST);
+    expect(Reflect.getMetadata('__guards__', ListController)).toEqual([FederationAuthGuard]);
+  });
+
+  it('runs AuthGuard context through ScopeService and returns local-source tagged rows', async () => {
+    const { controller, scope, query } = makeController();
+
+    const response = await controller.list('tasks', makeRequest(), { limit: 10 });
+
+    expect(scope.evaluateAccess).toHaveBeenCalledWith({
+      context: FEDERATION_CONTEXT,
+      resource: 'tasks',
+      requestedLimit: 10,
+      nativeRbac: query,
+    });
+    expect(query.list).toHaveBeenCalledWith({ filter: TASK_FILTER, cursor: undefined });
+    expect(response).toEqual({
+      items: [
+        {
+          id: 'task-1',
+          title: 'Federated task',
+          createdAt: new Date('2026-06-24T00:00:00.000Z'),
+          _source: 'local',
+        },
+      ],
+    });
+  });
+
+  it('preserves pagination metadata when row cap truncates the query layer result', async () => {
+    const { controller } = makeController({
+      queryResult: {
+        items: [{ id: 'task-1' }],
+        nextCursor: 'cursor-2',
+        truncated: true,
+      },
+    });
+
+    const response = await controller.list('tasks', makeRequest(), { cursor: 'cursor-1' });
+
+    expect(response).toEqual({
+      items: [{ id: 'task-1', _source: 'local' }],
+      nextCursor: 'cursor-2',
+      _truncated: true,
+    });
+  });
+
+  it('returns a federation error envelope when auth guard context is missing', async () => {
+    const { controller, scope, query } = makeController();
+
+    await expect(
+      controller.list('tasks', {} as unknown as FastifyRequest, {}),
+    ).rejects.toMatchObject({
+      response: {
+        error: {
+          code: 'unauthorized',
+          message: 'Federation context missing',
+        },
+      },
+      status: 401,
+    });
+    expect(scope.evaluateAccess).not.toHaveBeenCalled();
+    expect(query.list).not.toHaveBeenCalled();
+  });
+
+  it('returns a federation error envelope when scope evaluation denies access', async () => {
+    const { controller, query } = makeController({
+      scopeResult: {
+        allowed: false,
+        deny: {
+          code: 'resource_excluded',
+          stage: 'resource_exclusion',
+          statusCode: 403,
+          message: 'Requested federation resource is explicitly excluded by grant scope',
+          grantId: 'grant-1',
+          peerId: 'peer-1',
+          subjectUserId: 'user-1',
+          resource: 'credentials',
+        },
+      },
+    });
+
+    await expect(controller.list('credentials', makeRequest(), {})).rejects.toMatchObject({
+      response: {
+        error: {
+          code: 'scope_violation',
+          message: 'Requested federation resource is explicitly excluded by grant scope',
+        },
+      },
+      status: 403,
+    });
+    expect(query.list).not.toHaveBeenCalled();
+  });
+
+  it('rejects malformed request body fields before querying storage', async () => {
+    const { controller, scope, query } = makeController();
+
+    await expect(controller.list('tasks', makeRequest(), { cursor: 123 })).rejects.toMatchObject({
+      response: { error: { code: 'invalid_request' } },
+      status: 400,
+    });
+    await expect(controller.list('tasks', makeRequest(), { limit: false })).rejects.toMatchObject({
+      response: { error: { code: 'invalid_request' } },
+      status: 400,
+    });
+    await expect(controller.list('tasks', makeRequest(), { limit: 'abc' })).rejects.toMatchObject({
+      response: { error: { code: 'invalid_request' } },
+      status: 400,
+    });
+    expect(scope.evaluateAccess).not.toHaveBeenCalled();
+    expect(query.list).not.toHaveBeenCalled();
+  });
+});

apps/gateway/src/federation/server/verbs/capabilities.controller.ts

@@ -0,0 +1,38 @@
+/**
+ * Federation capabilities verb (FED-M3-07).
+ *
+ * Returns the read-only capability envelope for the active grant attached by
+ * FederationAuthGuard. This endpoint intentionally does not invoke native RBAC
+ * or ScopeService: an active grant is enough to ask what the grant allows.
+ */
+
+import { Controller, Get, Req, UseGuards } from '@nestjs/common';
+import type { FastifyRequest } from 'fastify';
+import {
+  FEDERATION_VERBS,
+  type FederationCapabilitiesResponse,
+  type FederationVerb,
+} from '@mosaicstack/types';
+import { parseFederationScope } from '../../scope-schema.js';
+import { FederationAuthGuard } from '../federation-auth.guard.js';
+import '../federation-context.js';
+
+@Controller('api/federation/v1/capabilities')
+@UseGuards(FederationAuthGuard)
+export class CapabilitiesController {
+  @Get()
+  getCapabilities(@Req() request: FastifyRequest): FederationCapabilitiesResponse {
+    if (!request.federationContext) {
+      throw new Error('Federation context missing after auth guard');
+    }
+
+    const scope = parseFederationScope(request.federationContext.scope);
+
+    return {
+      resources: [...scope.resources],
+      excluded_resources: [...scope.excluded_resources],
+      max_rows_per_query: scope.max_rows_per_query,
+      supported_verbs: [...FEDERATION_VERBS] satisfies FederationVerb[],
+    };
+  }
+}

apps/gateway/src/federation/server/verbs/list-query.service.ts

@@ -0,0 +1,408 @@
+/**
+ * Federation list query layer (FED-M3-05).
+ *
+ * Read-only DB adapter used by ListController after FederationAuthGuard and
+ * FederationScopeService have established the subject user, allowed resource,
+ * native-RBAC intersection, and row cap. Audit writes are intentionally
+ * deferred to M4.
+ */
+
+import { Inject, Injectable } from '@nestjs/common';
+import {
+  and,
+  desc,
+  eq,
+  inArray,
+  insights,
+  isNotNull,
+  lt,
+  missionTasks,
+  missions,
+  or,
+  preferences,
+  projects,
+  tasks,
+  teamMembers,
+  type Db,
+} from '@mosaicstack/db';
+import type {
+  FederationNativeRbacEvaluator,
+  FederationNativeRbacRequest,
+  FederationNativeRbacResult,
+  FederationScopeQueryFilter,
+} from '../scope.service.js';
+import { DB } from '../../../database/database.module.js';
+
+export interface FederationListQueryRequest {
+  readonly filter: FederationScopeQueryFilter;
+  readonly cursor?: string;
+}
+
+export interface FederationListQueryResult<T extends object = Record<string, unknown>> {
+  readonly items: T[];
+  readonly nextCursor?: string;
+  readonly truncated: boolean;
+}
+
+type CursorSource = 'insights' | 'preferences';
+const CURSOR_SOURCE = Symbol('federationCursorSource');
+
+type RowObject = Record<string, unknown> & { readonly [CURSOR_SOURCE]?: CursorSource };
+
+interface KeysetCursor {
+  readonly createdAt: Date;
+  readonly id: string;
+  readonly source?: CursorSource;
+}
+
+function encodeCursor(row: RowObject): string {
+  const createdAt = row['createdAt'];
+  const id = row['id'];
+  if (!(createdAt instanceof Date) || typeof id !== 'string') {
+    throw new Error('Federation list cursor cannot be encoded');
+  }
+
+  const source = row[CURSOR_SOURCE];
+  return Buffer.from(
+    JSON.stringify({ createdAt: createdAt.toISOString(), id, ...(source ? { source } : {}) }),
+    'utf8',
+  ).toString('base64url');
+}
+
+function decodeCursor(cursor: string | undefined): KeysetCursor | undefined {
+  if (cursor === undefined) {
+    return undefined;
+  }
+
+  try {
+    const parsed = JSON.parse(Buffer.from(cursor, 'base64url').toString('utf8')) as unknown;
+    if (typeof parsed !== 'object' || parsed === null) {
+      throw new Error('cursor must be an object');
+    }
+
+    const { createdAt, id, source } = parsed as {
+      createdAt?: unknown;
+      id?: unknown;
+      source?: unknown;
+    };
+    if (typeof createdAt !== 'string' || typeof id !== 'string' || id.length === 0) {
+      throw new Error('cursor is missing createdAt or id');
+    }
+    if (source !== undefined && source !== 'insights' && source !== 'preferences') {
+      throw new Error('cursor source is invalid');
+    }
+
+    const date = new Date(createdAt);
+    if (Number.isNaN(date.getTime())) {
+      throw new Error('cursor createdAt is invalid');
+    }
+
+    return { createdAt: date, id, ...(source ? { source } : {}) };
+  } catch {
+    throw new Error('Invalid federation list cursor');
+  }
+}
+
+function paginate<T extends RowObject>(rows: T[], limit: number): FederationListQueryResult<T> {
+  const page = rows.slice(0, limit);
+  const hasMore = rows.length > limit;
+  const nextCursor = hasMore ? encodeCursor(page[page.length - 1] ?? {}) : undefined;
+
+  return {
+    items: page,
+    truncated: hasMore,
+    ...(nextCursor !== undefined ? { nextCursor } : {}),
+  };
+}
+
+function markCursorSource<T extends RowObject>(row: T, source: CursorSource): T {
+  Object.defineProperty(row, CURSOR_SOURCE, {
+    value: source,
+    enumerable: false,
+    configurable: false,
+  });
+  return row;
+}
+
+function sortRows(rows: RowObject[]): RowObject[] {
+  return [...rows].sort((a, b) => {
+    const aTime = a['createdAt'] instanceof Date ? a['createdAt'].getTime() : 0;
+    const bTime = b['createdAt'] instanceof Date ? b['createdAt'].getTime() : 0;
+    if (aTime !== bTime) {
+      return bTime - aTime;
+    }
+    return String(b['id'] ?? '').localeCompare(String(a['id'] ?? ''));
+  });
+}
+
+@Injectable()
+export class FederationListQueryService implements FederationNativeRbacEvaluator {
+  constructor(@Inject(DB) private readonly db: Db) {}
+
+  async evaluateReadAccess(
+    request: FederationNativeRbacRequest,
+  ): Promise<FederationNativeRbacResult> {
+    if (request.resource === 'credentials' || request.resource === 'api_keys') {
+      return {
+        allowed: false,
+        reason: `${request.resource} federation list access is not implemented in M3`,
+        details: { resource: request.resource },
+      };
+    }
+
+    if (request.resource === 'memory') {
+      return { allowed: true, access: { includePersonal: true, teamIds: [] } };
+    }
+
+    const teamIds = await this.listSubjectTeamIds(request.subjectUserId);
+    return { allowed: true, access: { includePersonal: true, teamIds } };
+  }
+
+  async list<T extends RowObject = RowObject>(
+    request: FederationListQueryRequest,
+  ): Promise<FederationListQueryResult<T>> {
+    const cursor = decodeCursor(request.cursor);
+    const rows = await this.listAllRows(request.filter, request.filter.limit + 1, cursor);
+    return paginate(rows as T[], request.filter.limit);
+  }
+
+  private async listAllRows(
+    filter: FederationScopeQueryFilter,
+    rowLimit: number,
+    cursor: KeysetCursor | undefined,
+  ): Promise<RowObject[]> {
+    switch (filter.resource) {
+      case 'tasks':
+        return this.listTasks(filter, rowLimit, cursor);
+      case 'notes':
+        return this.listNotes(filter, rowLimit, cursor);
+      case 'memory':
+        return this.listMemory(filter, rowLimit, cursor);
+      case 'credentials':
+      case 'api_keys':
+        return [];
+      default:
+        throw new Error(`Unsupported federation list resource: ${String(filter.resource)}`);
+    }
+  }
+
+  private async listSubjectTeamIds(subjectUserId: string): Promise<string[]> {
+    const rows = await this.db
+      .select({ teamId: teamMembers.teamId })
+      .from(teamMembers)
+      .where(eq(teamMembers.userId, subjectUserId));
+
+    return rows.map((row) => row.teamId);
+  }
+
+  private async listAccessibleProjectIds(filter: FederationScopeQueryFilter): Promise<string[]> {
+    const clauses = [];
+    if (filter.includePersonal) {
+      clauses.push(and(eq(projects.ownerType, 'user'), eq(projects.ownerId, filter.subjectUserId)));
+    }
+    if (filter.teamIds.length > 0) {
+      clauses.push(
+        and(eq(projects.ownerType, 'team'), inArray(projects.teamId, [...filter.teamIds])),
+      );
+    }
+
+    if (clauses.length === 0) {
+      return [];
+    }
+
+    const rows = await this.db
+      .select({ id: projects.id })
+      .from(projects)
+      .where(clauses.length === 1 ? clauses[0] : or(...clauses));
+
+    return rows.map((row) => row.id);
+  }
+
+  private async listMissionIds(projectIds: readonly string[]): Promise<string[]> {
+    if (projectIds.length === 0) {
+      return [];
+    }
+
+    const rows = await this.db
+      .select({ id: missions.id })
+      .from(missions)
+      .where(inArray(missions.projectId, [...projectIds]));
+
+    return rows.map((row) => row.id);
+  }
+
+  private async listTasks(
+    filter: FederationScopeQueryFilter,
+    rowLimit: number,
+    cursor: KeysetCursor | undefined,
+  ): Promise<RowObject[]> {
+    const projectIds = await this.listAccessibleProjectIds(filter);
+    const missionIds = await this.listMissionIds(projectIds);
+    const clauses = [];
+
+    if (projectIds.length > 0) {
+      clauses.push(inArray(tasks.projectId, projectIds));
+    }
+    if (missionIds.length > 0) {
+      clauses.push(inArray(tasks.missionId, missionIds));
+    }
+
+    if (clauses.length === 0) {
+      return [];
+    }
+
+    const scopeClause = clauses.length === 1 ? clauses[0] : or(...clauses);
+    const cursorClause = cursor
+      ? or(
+          lt(tasks.createdAt, cursor.createdAt),
+          and(eq(tasks.createdAt, cursor.createdAt), lt(tasks.id, cursor.id)),
+        )
+      : undefined;
+
+    const rows = await this.db
+      .select({
+        id: tasks.id,
+        title: tasks.title,
+        description: tasks.description,
+        status: tasks.status,
+        priority: tasks.priority,
+        projectId: tasks.projectId,
+        missionId: tasks.missionId,
+        assignee: tasks.assignee,
+        tags: tasks.tags,
+        dueDate: tasks.dueDate,
+        metadata: tasks.metadata,
+        createdAt: tasks.createdAt,
+        updatedAt: tasks.updatedAt,
+      })
+      .from(tasks)
+      .where(and(scopeClause, cursorClause))
+      .orderBy(desc(tasks.createdAt), desc(tasks.id))
+      .limit(rowLimit);
+
+    return sortRows(rows as RowObject[]);
+  }
+
+  private async listNotes(
+    filter: FederationScopeQueryFilter,
+    rowLimit: number,
+    cursor: KeysetCursor | undefined,
+  ): Promise<RowObject[]> {
+    const projectIds = await this.listAccessibleProjectIds(filter);
+    const missionIds = await this.listMissionIds(projectIds);
+
+    if (missionIds.length === 0) {
+      return [];
+    }
+
+    // mission_tasks rows are user-scoped even when the mission belongs to a team.
+    // Team visibility can narrow the mission set, but it must never widen the
+    // query to other users' mission task notes.
+    const scopeClause = and(
+      eq(missionTasks.userId, filter.subjectUserId),
+      inArray(missionTasks.missionId, missionIds),
+    );
+    const cursorClause = cursor
+      ? or(
+          lt(missionTasks.createdAt, cursor.createdAt),
+          and(eq(missionTasks.createdAt, cursor.createdAt), lt(missionTasks.id, cursor.id)),
+        )
+      : undefined;
+
+    const rows = await this.db
+      .select({
+        id: missionTasks.id,
+        missionId: missionTasks.missionId,
+        taskId: missionTasks.taskId,
+        status: missionTasks.status,
+        content: missionTasks.notes,
+        createdAt: missionTasks.createdAt,
+        updatedAt: missionTasks.updatedAt,
+      })
+      .from(missionTasks)
+      .where(and(scopeClause, cursorClause, isNotNull(missionTasks.notes)))
+      .orderBy(desc(missionTasks.createdAt), desc(missionTasks.id))
+      .limit(rowLimit);
+
+    return sortRows(rows.filter((row) => row.content !== '') as RowObject[]);
+  }
+
+  private async listMemory(
+    filter: FederationScopeQueryFilter,
+    rowLimit: number,
+    cursor: KeysetCursor | undefined,
+  ): Promise<RowObject[]> {
+    if (!filter.includePersonal) {
+      return [];
+    }
+    if (cursor && cursor.source === undefined) {
+      throw new Error('Invalid federation list cursor');
+    }
+
+    const rows: RowObject[] = [];
+
+    // Memory spans two physical tables. To keep pagination deterministic and
+    // resumable without a SQL UNION, M3 emits a fixed block order: all insights
+    // first, then preferences. The opaque cursor records which table produced
+    // the boundary row, so the next page never re-applies one table's keyset to
+    // the other table (which could duplicate/skip rows at equal timestamps).
+    if (cursor?.source !== 'preferences') {
+      const insightCursorClause = cursor
+        ? or(
+            lt(insights.createdAt, cursor.createdAt),
+            and(eq(insights.createdAt, cursor.createdAt), lt(insights.id, cursor.id)),
+          )
+        : undefined;
+      const insightRows = await this.db
+        .select({
+          id: insights.id,
+          kind: insights.source,
+          content: insights.content,
+          category: insights.category,
+          relevanceScore: insights.relevanceScore,
+          metadata: insights.metadata,
+          createdAt: insights.createdAt,
+          updatedAt: insights.updatedAt,
+        })
+        .from(insights)
+        .where(and(eq(insights.userId, filter.subjectUserId), insightCursorClause))
+        .orderBy(desc(insights.createdAt), desc(insights.id))
+        .limit(rowLimit);
+
+      rows.push(...(insightRows as RowObject[]).map((row) => markCursorSource(row, 'insights')));
+    }
+
+    const remaining = rowLimit - rows.length;
+    if (remaining <= 0) {
+      return rows;
+    }
+
+    const preferenceCursorClause =
+      cursor?.source === 'preferences'
+        ? or(
+            lt(preferences.createdAt, cursor.createdAt),
+            and(eq(preferences.createdAt, cursor.createdAt), lt(preferences.id, cursor.id)),
+          )
+        : undefined;
+    const preferenceRows = await this.db
+      .select({
+        id: preferences.id,
+        kind: preferences.category,
+        key: preferences.key,
+        value: preferences.value,
+        source: preferences.source,
+        mutable: preferences.mutable,
+        createdAt: preferences.createdAt,
+        updatedAt: preferences.updatedAt,
+      })
+      .from(preferences)
+      .where(and(eq(preferences.userId, filter.subjectUserId), preferenceCursorClause))
+      .orderBy(desc(preferences.createdAt), desc(preferences.id))
+      .limit(remaining);
+
+    rows.push(
+      ...(preferenceRows as RowObject[]).map((row) => markCursorSource(row, 'preferences')),
+    );
+    return rows;
+  }
+}

apps/gateway/src/federation/server/verbs/list.controller.ts

@@ -0,0 +1,147 @@
+/**
+ * Federation list verb (FED-M3-05).
+ *
+ * POST /api/federation/v1/list/:resource
+ *
+ * Pipeline: FederationAuthGuard attaches the active grant context, then
+ * FederationScopeService enforces grant scope + native RBAC intersection, then
+ * the read-only query layer returns capped rows tagged with `_source`. Read
+ * audit-log writes are deferred to M4; this controller does not persist request
+ * or response bodies.
+ */
+
+import {
+  Body,
+  Controller,
+  HttpException,
+  Inject,
+  Param,
+  Post,
+  Req,
+  UseGuards,
+} from '@nestjs/common';
+import type { FastifyRequest } from 'fastify';
+import {
+  FederationInvalidRequestError,
+  FederationScopeViolationError,
+  FederationUnauthorizedError,
+  SOURCE_LOCAL,
+  tagWithSource,
+  type FederationListResponse,
+  type SourceTag,
+} from '@mosaicstack/types';
+import { FederationAuthGuard } from '../federation-auth.guard.js';
+import '../federation-context.js';
+import { FederationScopeService } from '../scope.service.js';
+import { FederationListQueryService } from './list-query.service.js';
+
+interface FederationListRequestBody {
+  readonly limit?: unknown;
+  readonly cursor?: unknown;
+}
+
+type FederatedRow = Record<string, unknown> & SourceTag;
+
+function parseLimit(body: FederationListRequestBody | undefined): number | undefined {
+  if (body?.limit === undefined) {
+    return undefined;
+  }
+
+  const parsed =
+    typeof body.limit === 'number'
+      ? body.limit
+      : typeof body.limit === 'string' && body.limit.trim().length > 0
+        ? Number(body.limit)
+        : Number.NaN;
+
+  if (!Number.isSafeInteger(parsed) || parsed < 1) {
+    throw new HttpException(
+      new FederationInvalidRequestError(
+        'Federation list limit must be a positive integer',
+      ).toEnvelope(),
+      400,
+    );
+  }
+
+  return parsed;
+}
+
+function parseCursor(body: FederationListRequestBody | undefined): string | undefined {
+  if (body?.cursor === undefined) {
+    return undefined;
+  }
+  if (typeof body.cursor === 'string') {
+    return body.cursor;
+  }
+  throw new HttpException(
+    new FederationInvalidRequestError('Federation list cursor must be a string').toEnvelope(),
+    400,
+  );
+}
+
+@Controller('api/federation/v1/list')
+@UseGuards(FederationAuthGuard)
+export class ListController {
+  constructor(
+    @Inject(FederationScopeService) private readonly scope: FederationScopeService,
+    @Inject(FederationListQueryService) private readonly query: FederationListQueryService,
+  ) {}
+
+  @Post(':resource')
+  async list(
+    @Param('resource') resource: string,
+    @Req() request: FastifyRequest,
+    @Body() body?: FederationListRequestBody,
+  ): Promise<FederationListResponse<FederatedRow>> {
+    if (!request.federationContext) {
+      throw new HttpException(
+        new FederationUnauthorizedError('Federation context missing').toEnvelope(),
+        401,
+      );
+    }
+
+    const requestedLimit = parseLimit(body);
+    const cursor = parseCursor(body);
+    const scopeResult = await this.scope.evaluateAccess({
+      context: request.federationContext,
+      resource,
+      requestedLimit,
+      nativeRbac: this.query,
+    });
+
+    if (!scopeResult.allowed) {
+      const ErrorClass =
+        scopeResult.deny.statusCode === 400
+          ? FederationInvalidRequestError
+          : FederationScopeViolationError;
+      throw new HttpException(
+        new ErrorClass(scopeResult.deny.message, scopeResult.deny).toEnvelope(),
+        scopeResult.deny.statusCode,
+      );
+    }
+
+    let result: Awaited<ReturnType<FederationListQueryService['list']>>;
+    try {
+      result = await this.query.list({ filter: scopeResult.filter, cursor });
+    } catch (error: unknown) {
+      if (error instanceof Error && error.message === 'Invalid federation list cursor') {
+        throw new HttpException(
+          new FederationInvalidRequestError('Federation list cursor is invalid').toEnvelope(),
+          400,
+        );
+      }
+      throw error;
+    }
+
+    const response: FederationListResponse<FederatedRow> = {
+      items: tagWithSource(result.items, SOURCE_LOCAL),
+    };
+    if (result.nextCursor !== undefined) {
+      response.nextCursor = result.nextCursor;
+    }
+    if (result.truncated) {
+      response._truncated = true;
+    }
+    return response;
+  }
+}

apps/gateway/src/plugin/discord-ingress.security.spec.ts

@@ -0,0 +1,89 @@
+import { describe, expect, it } from 'vitest';
+import {
+  createDiscordIngressEnvelope,
+  verifyDiscordIngressEnvelope,
+  type DiscordIngressPayload,
+} from '@mosaicstack/discord-plugin';
+import { validateDiscordServiceToken } from '../chat/chat.gateway-auth.js';
+import { DiscordReplayProtector } from './discord-replay-protector.js';
+
+const SERVICE_TOKEN = 'test-service-token';
+
+function createPayload(overrides: Partial<DiscordIngressPayload> = {}): DiscordIngressPayload {
+  return {
+    correlationId: 'correlation-001',
+    messageId: 'discord-message-001',
+    guildId: 'guild-001',
+    channelId: 'channel-001',
+    userId: 'user-001',
+    conversationId: 'discord-channel-001',
+    content: 'hello Tess',
+    ...overrides,
+  };
+}
+
+describe('Discord ingress security', () => {
+  it('accepts only the configured Discord service identity', () => {
+    expect(validateDiscordServiceToken(SERVICE_TOKEN, SERVICE_TOKEN)).toBe(true);
+    expect(validateDiscordServiceToken('wrong-service-token', SERVICE_TOKEN)).toBe(false);
+    expect(validateDiscordServiceToken(undefined, SERVICE_TOKEN)).toBe(false);
+  });
+
+  it('rejects unauthenticated or tampered service envelopes', () => {
+    const envelope = createDiscordIngressEnvelope(createPayload(), SERVICE_TOKEN);
+
+    expect(verifyDiscordIngressEnvelope(envelope, SERVICE_TOKEN)).toEqual(createPayload());
+    expect(verifyDiscordIngressEnvelope(envelope, 'wrong-service-token')).toBeNull();
+    expect(
+      verifyDiscordIngressEnvelope(
+        { ...envelope, payload: { ...envelope.payload, content: 'forged command' } },
+        SERVICE_TOKEN,
+      ),
+    ).toBeNull();
+  });
+
+  it.each([
+    ['guild', { guildId: 'unlisted-guild' }],
+    ['channel', { channelId: 'unlisted-channel' }],
+    ['user', { userId: 'unlisted-user' }],
+  ])(
+    'rejects an unallowlisted Discord %s',
+    (_kind: string, overrides: Partial<DiscordIngressPayload>) => {
+      const envelope = createDiscordIngressEnvelope(createPayload(overrides), SERVICE_TOKEN);
+
+      expect(
+        verifyDiscordIngressEnvelope(envelope, SERVICE_TOKEN, {
+          guildIds: ['guild-001'],
+          channelIds: ['channel-001'],
+          userIds: ['user-001'],
+        }),
+      ).toBeNull();
+    },
+  );
+
+  it('retains Discord message and correlation IDs after authenticated allowlisted validation', () => {
+    const payload = createPayload({
+      correlationId: 'correlation-trace-123',
+      messageId: 'discord-snowflake-987',
+    });
+    const envelope = createDiscordIngressEnvelope(payload, SERVICE_TOKEN);
+
+    expect(
+      verifyDiscordIngressEnvelope(envelope, SERVICE_TOKEN, {
+        guildIds: ['guild-001'],
+        channelIds: ['channel-001'],
+        userIds: ['user-001'],
+      }),
+    ).toEqual(payload);
+  });
+
+  it('rejects a replayed Discord message ID while retaining bounded replay state', () => {
+    const replayProtector = new DiscordReplayProtector(60_000, 2);
+
+    expect(replayProtector.claim('discord-message-001')).toBe(true);
+    expect(replayProtector.claim('discord-message-001')).toBe(false);
+    expect(replayProtector.claim('discord-message-002')).toBe(true);
+    expect(replayProtector.claim('discord-message-003')).toBe(true);
+    expect(replayProtector.size).toBe(2);
+  });
+});

apps/gateway/src/plugin/discord-replay-protector.ts

@@ -0,0 +1,40 @@
+/**
+ * Bounded replay cache for Discord's globally unique native message IDs.
+ * Durable ingress idempotency is added with Tess's canonical inbox/outbox work.
+ */
+export class DiscordReplayProtector {
+  private readonly claimedAt = new Map<string, number>();
+
+  constructor(
+    private readonly ttlMs = 15 * 60 * 1000,
+    private readonly maxEntries = 10_000,
+  ) {}
+
+  get size(): number {
+    return this.claimedAt.size;
+  }
+
+  /** Claims an ID exactly once within its bounded retention window. */
+  claim(messageId: string, now = Date.now()): boolean {
+    this.prune(now);
+    if (this.claimedAt.has(messageId)) return false;
+
+    this.claimedAt.set(messageId, now);
+    this.evictOverflow();
+    return true;
+  }
+
+  private prune(now: number): void {
+    for (const [messageId, claimedAt] of this.claimedAt) {
+      if (now - claimedAt >= this.ttlMs) this.claimedAt.delete(messageId);
+    }
+  }
+
+  private evictOverflow(): void {
+    while (this.claimedAt.size > this.maxEntries) {
+      const oldestMessageId = this.claimedAt.keys().next().value;
+      if (oldestMessageId === undefined) return;
+      this.claimedAt.delete(oldestMessageId);
+    }
+  }
+}

apps/gateway/src/plugin/plugin.module.ts

@@ -50,19 +50,41 @@ class TelegramChannelPluginAdapter implements IChannelPlugin {
 
 const DEFAULT_GATEWAY_URL = 'http://localhost:14242';
 
+function requiredDiscordAllowlist(name: string): string[] {
+  const value = process.env[name]
+    ?.split(',')
+    .map((id: string): string => id.trim())
+    .filter((id: string): boolean => id.length > 0);
+  if (!value || value.length === 0) {
+    throw new Error(`${name} is required when DISCORD_BOT_TOKEN is configured`);
+  }
+  return value;
+}
+
 function createPluginRegistry(): IChannelPlugin[] {
   const plugins: IChannelPlugin[] = [];
   const discordToken = process.env['DISCORD_BOT_TOKEN'];
   const discordGuildId = process.env['DISCORD_GUILD_ID'];
   const discordGatewayUrl = process.env['DISCORD_GATEWAY_URL'] ?? DEFAULT_GATEWAY_URL;
+  const discordServiceToken = process.env['DISCORD_SERVICE_TOKEN'];
+  const discordServiceUserId = process.env['DISCORD_SERVICE_USER_ID'];
 
   if (discordToken) {
+    if (!discordServiceToken || !discordServiceUserId) {
+      throw new Error(
+        'DISCORD_SERVICE_TOKEN and DISCORD_SERVICE_USER_ID are required when DISCORD_BOT_TOKEN is configured',
+      );
+    }
     plugins.push(
       new DiscordChannelPluginAdapter(
         new DiscordPlugin({
           token: discordToken,
           guildId: discordGuildId,
           gatewayUrl: discordGatewayUrl,
+          serviceToken: discordServiceToken,
+          allowedGuildIds: requiredDiscordAllowlist('DISCORD_ALLOWED_GUILD_IDS'),
+          allowedChannelIds: requiredDiscordAllowlist('DISCORD_ALLOWED_CHANNEL_IDS'),
+          allowedUserIds: requiredDiscordAllowlist('DISCORD_ALLOWED_USER_IDS'),
         }),
       ),
     );

docs/MISSION-MANIFEST.md

@@ -67,10 +67,11 @@ The MVP is complete when ALL declared workstreams are complete AND every cross-c
 
 ## Workstreams
 
-| #   | ID  | Name                                        | Status            | Manifest                                                                | Notes                                               |
-| --- | --- | ------------------------------------------- | ----------------- | ----------------------------------------------------------------------- | --------------------------------------------------- |
-| W1  | FED | Federation v1                               | planning-complete | [docs/federation/MISSION-MANIFEST.md](./federation/MISSION-MANIFEST.md) | 7 milestones, ~175K tokens, issues #460–#466 filed  |
-| W2+ | TBD | (additional workstreams declared as scoped) | —                 | —                                                                       | Scope creep is expected and explicitly accommodated |
+| #   | ID   | Name                                        | Status            | Manifest                                                                | Notes                                               |
+| --- | ---- | ------------------------------------------- | ----------------- | ----------------------------------------------------------------------- | --------------------------------------------------- |
+| W1  | FED  | Federation v1                               | planning-complete | [docs/federation/MISSION-MANIFEST.md](./federation/MISSION-MANIFEST.md) | 7 milestones, ~175K tokens, issues #460–#466 filed  |
+| W2  | TESS | Tess interaction agent                      | planning-complete | [docs/tess/MISSION-MANIFEST.md](./tess/MISSION-MANIFEST.md)             | 5 milestones; issue #706; M1 issue #707 ready       |
+| W3+ | TBD  | (additional workstreams declared as scoped) | —                 | —                                                                       | Scope creep is expected and explicitly accommodated |
 
 ### Likely Additional Workstreams (Not Yet Declared)
 

docs/PRD.md

@@ -79,6 +79,102 @@ Jarvis (v0.2.0) is a self-hosted AI assistant with a Python FastAPI backend and
 
 ---
 
+## Tess Interaction Agent Workstream (TESS)
+
+### Problem and Objective
+
+Jason needs one durable, operator-facing Mosaic agent outside Hermes that is reachable through a dedicated Discord channel and CLI, can attach to and operate the Mosaic fleet and transitional Hermes agents, and preserves context across restarts and compaction. Mos remains the coding/general fleet orchestrator; Tess is the complementary human interaction, visibility, control, and migration agent.
+
+The objective is to ship **Tess** (from _tessera_, a piece of a mosaic) as a Pi-native, GPT-5.6 Sol agent with high reasoning. Tess must use Mosaic-owned contracts and plugins so Hermes can be replaced incrementally rather than becoming a permanent architectural dependency.
+
+### Scope
+
+#### In Scope
+
+1. `TESS-ARP-001`: A runtime-neutral `AgentRuntimeProvider` contract supporting `listSessions`, `streamSession`, `sendMessage`, `terminate`, `getSessionTree`, `attach`, health, capability discovery, and normalized events/errors.
+2. `TESS-PI-001`: A long-running Pi-native Tess agent profile/service pinned to GPT-5.6 Sol with high reasoning, explicit tool policy, lifecycle hooks, durable checkpoints, and restart recovery.
+3. `TESS-DSC-001`: Dedicated Discord channel binding to Tess through the Mosaic gateway, with allowlists/RBAC, thread/reply policy, streaming, attachments, approvals, and correlation IDs.
+4. `TESS-CLI-001`: `mosaic tess` CLI commands for chat, status, session listing, attach/detach, send/steer/stop, provider health, and recovery.
+5. `TESS-FLT-001`: Fleet plugin capabilities for roster/status/heartbeat inspection, message delivery, session hierarchy, safe attach, and controlled restart/recovery.
+6. `TESS-MOS-001`: Explicit Mos coordination boundary and tools: hand off orchestration requests, observe mission/task state, receive results, and never silently compete for orchestration authority.
+7. `TESS-HRM-001`: Transitional Hermes adapter for profiles/agents, sessions, streaming/messages, Kanban, skills, memory, tools, cron, and health, using capability negotiation and fail-closed unsupported operations.
+8. `TESS-MEM-001`: Unified memory/retrieval plugin with scoped search/recent/capture/stats, startup context injection, provenance, redaction, namespace isolation, and flat-file/project truth precedence.
+9. `TESS-STA-001`: Durable agent state, inbox, handoff, compaction-recovery, and resume reconstruction.
+10. `TESS-PLG-001`: Plugin/tool catalog covering runtime bootstrap, repository/PR workflow, fleet diagnostics, incident-safe read operations, Discord interaction, and extensible MCP/skill discovery.
+11. `TESS-TRN-001`: Replaceable transport providers: tmux/fleet now, Matrix/native Mosaic transport later, with no Discord/CLI business logic coupled to transport details.
+12. `TESS-SEC-001`: RBAC, per-operation authorization, explicit approval for destructive/privileged/customer-visible actions, audit events, secret/PII redaction, tenant isolation, and bounded command execution.
+13. `TESS-SEC-002`: Command execution SHALL enforce declared scope/role server-side; admin/system and destructive operations SHALL require policy-bound durable approval.
+14. `TESS-SEC-003`: Every session list/read/attach/send/terminate operation SHALL enforce server-derived owner and tenant scope; guessed or client-supplied IDs SHALL grant no authority.
+15. `TESS-SEC-004`: MCP tools SHALL derive actor/tenant from authenticated context and SHALL NOT accept caller-controlled identity fields.
+16. `TESS-SEC-005`: Discord plugin ingress SHALL authenticate service identity, enforce guild/channel/user allowlists, propagate correlation/message IDs, and reject replay.
+17. `TESS-SEC-006`: Secret/PII classification and redaction SHALL occur before persistence and before channel egress, including tool metadata and authentication flows.
+18. `TESS-SEC-007`: Approvals SHALL be one-time, expiring, actor/tenant-bound, and cryptographically bound to the exact structured action digest.
+19. `TESS-SEC-008`: Ingress, provider sends, tool side effects, and responses SHALL use durable inbox/outbox/checkpoints and idempotency records for restart-safe replay.
+20. `TESS-SEC-009`: Garbage collection and retention SHALL be session/tenant scoped unless executed as a separately authorized and audited system-wide job.
+21. `TESS-OBS-001`: Structured logs, traces, health/readiness, provider latency/errors, session lifecycle, tool audit, and actionable recovery diagnostics.
+22. `TESS-MIG-001`: Capability inventory and staged Hermes-to-Mosaic migration matrix with coexistence, cutover, rollback, and deprecation gates.
+
+#### Out of Scope
+
+1. Replacing Mos as coding/general fleet orchestrator.
+2. Making Hermes the Mosaic core or coupling Mosaic domain logic to Hermes schemas.
+3. Migrating every historical chat verbatim; only policy-compliant indexed summaries and user-selected sessions are migrated.
+4. Unrestricted shell execution from Discord.
+5. Full web UI parity in the first Tess operational milestone; gateway contracts must remain web-consumable.
+6. Replacing tmux before Matrix/native transport reaches operational parity.
+
+### Stakeholder and User Requirements
+
+- Jason must be able to converse with the same Tess session from Discord and CLI.
+- Jason must be able to see what is running, stale, blocked, or unhealthy without attaching manually to every session.
+- Jason must be able to attach to Tess and authorized fleet sessions through supported CLI controls.
+- Tess must collaborate with Mos and the fleet while preserving a single clear orchestration authority.
+- The system must migrate useful Hermes/OpenClaw capabilities intentionally, with evidence, instead of copying implementations wholesale.
+
+### Non-Functional Requirements
+
+1. **Security:** default-deny provider/tool capabilities, least privilege, no secrets in logs/prompts/commits, Discord user/channel authorization, and auditable approvals.
+2. **Reliability:** durable inbox/checkpoints; idempotent message handling; reconnect with bounded backoff; no message loss or duplicate execution across gateway restart.
+3. **Performance:** first acknowledgement within 2 seconds when connected; streamed agent output begins within 5 seconds excluding model/provider delay; status reads return within 2 seconds under nominal local conditions.
+4. **Observability:** every ingress message and resulting provider/tool operation carries a correlation ID across Discord, gateway, Tess, provider, and audit events.
+5. **Maintainability:** channel, runtime, transport, memory, and external-agent integrations remain adapter-based with contract tests.
+6. **Privacy:** only scoped context enters external runtimes; persisted messages/memories follow retention and redaction policy.
+7. **Portability:** Tess runs through Pi/Mosaic contracts and does not require Hermes to start or serve native Mosaic operations.
+
+### Acceptance Criteria
+
+1. `AC-TESS-01`: A dedicated Discord channel and `mosaic tess chat` connect to one durable Tess session and stream responses bidirectionally.
+2. `AC-TESS-02`: `mosaic tess status|sessions|tree|attach|send|stop` operate against authorized provider capabilities with stable typed outputs and actionable errors.
+3. `AC-TESS-03`: Tess runs GPT-5.6 Sol at high reasoning and its effective runtime/model/tool policy is visible through status without exposing credentials.
+4. `AC-TESS-04`: Tess can inspect and message the Mosaic fleet, hand orchestration work to Mos, and demonstrate that Tess does not independently claim Mos-owned orchestration work.
+5. `AC-TESS-05`: Hermes adapter demonstrates session listing, streaming/message delivery, hierarchy mapping, and at least one approved capability in each of Kanban, skills, memory, tools, and cron—or reports unsupported capabilities fail-closed.
+6. `AC-TESS-06`: Restart/compaction test preserves session identity, pending inbox, last durable checkpoint, and a resumable handoff without duplicate side effects.
+7. `AC-TESS-07`: Unauthorized Discord users/channels, cross-tenant access, unsafe tool calls, forged approvals, and sensitive-output cases are denied and audited.
+8. `AC-TESS-08`: tmux/fleet and Matrix/native transport implementations pass the same provider contract suite; Matrix may remain non-default until readiness gates pass.
+9. `AC-TESS-09`: Baseline quality gates, unit/integration/contract tests, Discord+CLI E2E, restart/recovery tests, independent code review, and security review are green.
+10. `AC-TESS-10`: Migration matrix documents every audited Hermes/OpenClaw capability as native, adapted, deferred, or rejected, with cutover and rollback evidence.
+11. `AC-TESS-11`: User, admin, developer, API/OpenAPI, operations/recovery, and plugin-authoring documentation is current and linked from the sitemap.
+
+### Constraints, Dependencies, Risks, and Assumptions
+
+- Dependency: Mosaic gateway remains the single API surface; Pi is the native runtime; Valkey/PostgreSQL provide canonical durable state where required.
+- Dependency: Discord bot credentials and dedicated channel ID are deployment secrets provisioned outside source control.
+- Risk: Tess could drift into a second orchestrator. Mitigation: explicit role policy, Mos handoff contract, authority checks, and E2E boundary tests.
+- Risk: broad Hermes compatibility can freeze legacy semantics into Mosaic. Mitigation: Mosaic-owned normalized contracts and capability negotiation.
+- Risk: Discord creates a privileged remote-control surface. Mitigation: pairing/allowlists, RBAC, approvals, rate limits, audit, and safe tool classes.
+- Risk: transcript ingestion can violate privacy or overload memory. Mitigation: scoped opt-in import, redacted summaries, provenance, retention, and deduplication.
+- Risk: current root filesystem has limited headroom. Mitigation: isolated worktrees, no duplicated dependency installation unless required, and cleanup only after active-lane verification.
+- `ASSUMPTION:` The public name is **Tess**, because the user requested a name and the tessera/Mosaic relationship is distinctive; config must permit later display-name changes without renaming APIs or storage keys.
+- `ASSUMPTION:` The dedicated Discord channel ID and final guild policy will be supplied/provisioned during deployment, so implementation uses explicit configuration and fail-fast startup validation.
+- `ASSUMPTION:` tmux/fleet is the production transport for the first operational milestone; Matrix/native transport is implemented behind the same contract and promoted only after parity/reliability verification.
+- `ASSUMPTION:` Project/task truth remains in canonical Mosaic/project stores; semantic memory systems are retrieval/mirror layers, not hidden authorities.
+
+### Testing and Delivery Intent
+
+Delivery uses five gated milestones: runtime contracts/security; Pi service/state; Discord/CLI; fleet/Hermes/plugin suite; migration/Matrix/recovery/qualification. Every source-code task requires tests, independent review, a PR to `main`, terminal-green CI, and issue/task closure. Production activation additionally requires a clean-host Pi launch, dedicated Discord channel smoke test, CLI attach test, restart/recovery drill, and rollback procedure.
+
+---
+
 ## Architecture
 
 ### High-Level System Diagram

docs/TASKS.md

@@ -14,9 +14,10 @@
 
 ## Workstream Rollup
 
-| id  | status            | workstream          | progress         | tasks file                                        | notes                                                           |
-| --- | ----------------- | ------------------- | ---------------- | ------------------------------------------------- | --------------------------------------------------------------- |
-| W1  | planning-complete | Federation v1 (FED) | 0 / 7 milestones | [docs/federation/TASKS.md](./federation/TASKS.md) | M1 task breakdown populated; M2–M7 deferred to mission planning |
+| id  | status            | workstream             | progress         | tasks file                                        | notes                                                           |
+| --- | ----------------- | ---------------------- | ---------------- | ------------------------------------------------- | --------------------------------------------------------------- |
+| W1  | planning-complete | Federation v1 (FED)    | 0 / 7 milestones | [docs/federation/TASKS.md](./federation/TASKS.md) | M1 task breakdown populated; M2–M7 deferred to mission planning |
+| W2  | planning-complete | Tess interaction agent | 0 / 5 milestones | [docs/tess/TASKS.md](./tess/TASKS.md)             | Issue #706; independent planning gate PASS; M1 issue #707 ready |
 
 ## Cross-Cutting Tracking
 

docs/architecture/channel-protocol.md

@@ -232,6 +232,10 @@ The following sections document how each supported channel maps its native messa
 
 **Outbound:** Adapter calls Discord REST `POST /channels/{id}/messages`. Markdown content is sent as-is (Discord renders it). For `contentType = "code"` the adapter wraps in triple-backtick fences with the `metadata.language` tag.
 
+### Discord service ingress security
+
+The Discord adapter is an authenticated gateway service, not an anonymous Socket.IO client. It presents `DISCORD_SERVICE_TOKEN` during its `/chat` connection and signs each inbound envelope using HMAC-SHA-256. The envelope contains the Discord native message ID and a generated correlation ID. Gateway verifies the service credential, signature, and configured guild/channel/user allowlists before agent dispatch, then rejects duplicate native message IDs inside its bounded replay window. All three allowlists are default-deny and required when the Discord plugin is enabled. The service credential is injected at runtime and is never logged or included in protocol payloads.
+
 ---
 
 ### Telegram

docs/federation/TASKS.md

@@ -91,22 +91,22 @@ Goal: Two federated gateways exchange real data over mTLS. Inbound requests pass
 >
 > **Tracking issue:** #462.
 
-| id        | status      | description                                                                                                                                                                                                                                                                                            | issue | agent  | branch                               | depends_on       | estimate | notes                                                                                                                                                    |
-| --------- | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----- | ------ | ------------------------------------ | ---------------- | -------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| FED-M3-01 | not-started | `packages/types/src/federation/` — request/response DTOs for `list`, `get`, `capabilities` verbs. Wire-format zod schemas + inferred TS types. Includes `FederationRequest`, `FederationListResponse<T>`, `FederationGetResponse<T>`, `FederationCapabilitiesResponse`, error envelope, `_source` tag. | #462  | sonnet | feat/federation-m3-types             | —                | 4K       | Reusable from gateway server + client + harness. Pure types — no I/O, no NestJS.                                                                         |
-| FED-M3-02 | not-started | `tools/federation-harness/` scaffold: `docker-compose.two-gateways.yml` (Server A + Server B + step-CA), `seed.ts` (provisions grants, peers, sample tasks/notes/credentials per scope variant), `harness.ts` helper (boots stack, returns typed clients). README documents harness use.               | #462  | sonnet | feat/federation-m3-harness           | DEPLOY-04 (soft) | 8K       | Falls back to local docker-compose if `mos-test-1/-2` not yet redeployed (DEPLOY chain blocked on IMG-FIX). Permanent test infra used by M3+.            |
-| FED-M3-03 | not-started | `apps/gateway/src/federation/server/federation-auth.guard.ts` (NestJS guard). Validates inbound client cert from Fastify TLS context, extracts `grantId` + `subjectUserId` from custom OIDs, loads grant from DB, asserts `status='active'`, attaches `FederationContext` to request.                  | #462  | sonnet | feat/federation-m3-auth-guard        | M3-01            | 8K       | Reuses OID parsing logic mirrored from `ca.service.ts` post-issuance verification. 401 on malformed/missing OIDs; 403 on revoked/expired/missing grant.  |
-| FED-M3-04 | not-started | `apps/gateway/src/federation/server/scope.service.ts`. Pipeline: (1) resource allowlist + excluded check, (2) native RBAC eval as `subjectUserId`, (3) scope filter intersection (`include_teams`, `include_personal`), (4) `max_rows_per_query` cap. Pure service — DB calls injected.                | #462  | sonnet | feat/federation-m3-scope-service     | M3-01            | 10K      | Hardest correctness target in M3. Reuses `parseFederationScope` (M2-03). Returns either `{ allowed: true, filter }` or structured deny reason for audit. |
-| FED-M3-05 | not-started | `apps/gateway/src/federation/server/verbs/list.controller.ts`. Wires AuthGuard → ScopeService → tasks/notes/memory query layer; applies row cap; tags rows with `_source`. Resource selector via path param.                                                                                           | #462  | sonnet | feat/federation-m3-verb-list         | M3-03, M3-04     | 6K       | Routes: `POST /api/federation/v1/list/:resource`. No body persistence. Audit write deferred to M4.                                                       |
-| FED-M3-06 | not-started | `apps/gateway/src/federation/server/verbs/get.controller.ts`. Single-resource fetch by id; same pipeline as list. 404 on not-found, 403 on RBAC/scope deny — both audited the same way.                                                                                                                | #462  | sonnet | feat/federation-m3-verb-get          | M3-03, M3-04     | 6K       | `POST /api/federation/v1/get/:resource/:id`. Mirrors list controller patterns.                                                                           |
-| FED-M3-07 | not-started | `apps/gateway/src/federation/server/verbs/capabilities.controller.ts`. Read-only enumeration: returns `{ resources, excluded_resources, max_rows_per_query, supported_verbs }` derived from grant scope. Always allowed for an active grant — no RBAC eval.                                            | #462  | sonnet | feat/federation-m3-verb-capabilities | M3-03            | 4K       | `GET /api/federation/v1/capabilities`. Smallest verb; useful sanity check that mTLS + auth guard work end-to-end.                                        |
-| FED-M3-08 | not-started | `apps/gateway/src/federation/client/federation-client.service.ts`. Outbound mTLS dialer: picks `(certPem, sealed clientKey)` from `federation_peers`, unwraps key, builds undici Agent with mTLS, calls peer verb, parses typed response, wraps non-2xx into `FederationClientError`.                  | #462  | sonnet | feat/federation-m3-client            | M3-01            | 8K       | Independent of server stream — can land in parallel with M3-03/04. Cert/key cached per-peer; flushed by future M5/M6 logic.                              |
-| FED-M3-09 | not-started | `apps/gateway/src/federation/client/query-source.service.ts`. Accepts `source: "local" \| "federated:<host>" \| "all"` from gateway query layer; for `"all"` fans out to local + each peer in parallel; merges results; tags every row with `_source`.                                                 | #462  | sonnet | feat/federation-m3-query-source      | M3-08            | 8K       | Per-peer failure surfaces as `_partial: true` in response, not hard failure (sets up M5 offline UX). M5 adds caching + circuit breaker on top.           |
-| FED-M3-10 | not-started | Integration tests for MILESTONES.md M3 acceptance #6 (malformed OIDs → 401; valid cert + revoked grant → 403) and #7 (`max_rows_per_query` cap). Real PG, mocked TLS context (Fastify req shim).                                                                                                       | #462  | sonnet | feat/federation-m3-integration       | M3-05, M3-06     | 8K       | Vitest profile gated by `FEDERATED_INTEGRATION=1`. Single-gateway suite; no harness required.                                                            |
-| FED-M3-11 | not-started | E2E tests for MILESTONES.md M3 acceptance #1, #2, #3, #4, #5, #8, #9, #10 (8 cases). Uses harness from M3-02; two real gateways, real Step-CA, real mTLS. Each test asserts both happy-path response and audit/no-persist invariants.                                                                  | #462  | sonnet | feat/federation-m3-e2e               | M3-02, M3-09     | 12K      | Largest single task. Each acceptance gets its own `it(...)` for clear failure attribution.                                                               |
-| FED-M3-12 | not-started | Independent security review (sonnet, not author of M3-03/04/05/06/07/08/09): focus on cert-SAN spoofing, OID extraction edge cases, scope-bypass via filter manipulation, RBAC-bypass via subjectUser swap, response leakage when scope deny.                                                          | #462  | sonnet | feat/federation-m3-security-review   | M3-11            | 10K      | Two review rounds budgeted. PRD requires explicit test for every 401/403 path — review verifies coverage.                                                |
-| FED-M3-13 | not-started | Docs update: `docs/federation/SETUP.md` mTLS handshake section, new `docs/federation/HARNESS.md` for federation-harness usage, OID reference table in SETUP.md, scope enforcement pipeline diagram. Runbook still M7-deferred.                                                                         | #462  | haiku  | feat/federation-m3-docs              | M3-12            | 5K       | One ASCII diagram for the auth-guard → scope → RBAC pipeline; helps future reviewers reason about denial paths.                                          |
-| FED-M3-14 | not-started | PR aggregate close, CI green, merge to main, close #462. Release tag `fed-v0.3.0-m3`. Update mission manifest M3 row → done; M4 row → in-progress when work begins.                                                                                                                                    | #462  | sonnet | chore/federation-m3-close            | M3-13            | 3K       | Same close pattern as M1-12 / M2-13.                                                                                                                     |
+| id        | status      | description                                                                                                                                                                                                                                                                                            | issue | agent  | branch                               | depends_on                        | estimate | notes                                                                                                                                                    |
+| --------- | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----- | ------ | ------------------------------------ | --------------------------------- | -------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| FED-M3-01 | done        | `packages/types/src/federation/` — request/response DTOs for `list`, `get`, `capabilities` verbs. Wire-format zod schemas + inferred TS types. Includes `FederationRequest`, `FederationListResponse<T>`, `FederationGetResponse<T>`, `FederationCapabilitiesResponse`, error envelope, `_source` tag. | #462  | sonnet | feat/federation-m3-types             | —                                 | 4K       | Reusable from gateway server + client + harness. Pure types — no I/O, no NestJS.                                                                         |
+| FED-M3-02 | done        | `tools/federation-harness/` scaffold: `docker-compose.two-gateways.yml` (Server A + Server B + step-CA), `seed.ts` (provisions grants, peers, sample tasks/notes/credentials per scope variant), `harness.ts` helper (boots stack, returns typed clients). README documents harness use.               | #462  | sonnet | feat/federation-m3-harness           | DEPLOY-04 (soft)                  | 8K       | Falls back to local docker-compose if `mos-test-1/-2` not yet redeployed (DEPLOY chain blocked on IMG-FIX). Permanent test infra used by M3+.            |
+| FED-M3-03 | done        | `apps/gateway/src/federation/server/federation-auth.guard.ts` (NestJS guard). Validates inbound client cert from Fastify TLS context, extracts `grantId` + `subjectUserId` from custom OIDs, loads grant from DB, asserts `status='active'`, attaches `FederationContext` to request.                  | #462  | sonnet | feat/federation-m3-auth-guard        | M3-01                             | 8K       | Reuses OID parsing logic mirrored from `ca.service.ts` post-issuance verification. 401 on malformed/missing OIDs; 403 on revoked/expired/missing grant.  |
+| FED-M3-04 | in-progress | `apps/gateway/src/federation/server/scope.service.ts`. Pipeline: (1) resource allowlist + excluded check, (2) native RBAC eval as `subjectUserId`, (3) scope filter intersection (`include_teams`, `include_personal`), (4) `max_rows_per_query` cap. Pure service — DB calls injected.                | #462  | sonnet | feat/federation-m3-scope-service     | M3-01                             | 10K      | Hardest correctness target in M3. Reuses `parseFederationScope` (M2-03). Returns either `{ allowed: true, filter }` or structured deny reason for audit. |
+| FED-M3-05 | in-progress | `apps/gateway/src/federation/server/verbs/list.controller.ts`. Wires AuthGuard → ScopeService → tasks/notes/memory query layer; applies row cap; tags rows with `_source`. Resource selector via path param.                                                                                           | #462  | sonnet | feat/federation-m3-verb-list         | M3-03, M3-04                      | 6K       | Routes: `POST /api/federation/v1/list/:resource`. No body persistence. Audit write deferred to M4.                                                       |
+| FED-M3-06 | not-started | `apps/gateway/src/federation/server/verbs/get.controller.ts`. Single-resource fetch by id; same pipeline as list. 404 on not-found, 403 on RBAC/scope deny — both audited the same way.                                                                                                                | #462  | sonnet | feat/federation-m3-verb-get          | M3-03, M3-04                      | 6K       | `POST /api/federation/v1/get/:resource/:id`. Mirrors list controller patterns.                                                                           |
+| FED-M3-07 | done        | `apps/gateway/src/federation/server/verbs/capabilities.controller.ts`. Read-only enumeration: returns `{ resources, excluded_resources, max_rows_per_query, supported_verbs }` derived from grant scope. Always allowed for an active grant — no RBAC eval.                                            | #462  | sonnet | feat/federation-m3-verb-capabilities | M3-03                             | 4K       | `GET /api/federation/v1/capabilities`. Smallest verb; useful sanity check that mTLS + auth guard work end-to-end.                                        |
+| FED-M3-08 | done        | `apps/gateway/src/federation/client/federation-client.service.ts`. Outbound mTLS dialer: picks `(certPem, sealed clientKey)` from `federation_peers`, unwraps key, builds undici Agent with mTLS, calls peer verb, parses typed response, wraps non-2xx into `FederationClientError`.                  | #462  | sonnet | feat/federation-m3-client            | M3-01                             | 8K       | Independent of server stream — can land in parallel with M3-03/04. Cert/key cached per-peer; flushed by future M5/M6 logic.                              |
+| FED-M3-09 | done        | `apps/gateway/src/federation/client/query-source.service.ts`. Accepts `source: "local" \| "federated:<host>" \| "all"` from gateway query layer; for `"all"` fans out to local + each peer in parallel; merges results; tags every row with `_source`.                                                 | #462  | sonnet | feat/federation-m3-query-source      | M3-08                             | 8K       | Per-peer failure surfaces as `_partial: true` in response, not hard failure (sets up M5 offline UX). M5 adds caching + circuit breaker on top.           |
+| FED-M3-10 | not-started | Integration tests for MILESTONES.md M3 acceptance #6 (malformed OIDs → 401; valid cert + revoked grant → 403) and #7 (`max_rows_per_query` cap). Real PG, mocked TLS context (Fastify req shim).                                                                                                       | #462  | sonnet | feat/federation-m3-integration       | M3-05, M3-06                      | 8K       | Vitest profile gated by `FEDERATED_INTEGRATION=1`. Single-gateway suite; no harness required.                                                            |
+| FED-M3-11 | not-started | E2E tests for MILESTONES.md M3 acceptance #1, #2, #3, #4, #5, #8, #9, #10 (8 cases). Uses harness from M3-02; two real gateways, real Step-CA, real mTLS. Each test asserts both happy-path response and audit/no-persist invariants.                                                                  | #462  | sonnet | feat/federation-m3-e2e               | M3-02, M3-04, M3-05, M3-06, M3-09 | 12K      | Largest single task. Each acceptance gets its own `it(...)` for clear failure attribution.                                                               |
+| FED-M3-12 | not-started | Independent security review (sonnet, not author of M3-03/04/05/06/07/08/09): focus on cert-SAN spoofing, OID extraction edge cases, scope-bypass via filter manipulation, RBAC-bypass via subjectUser swap, response leakage when scope deny.                                                          | #462  | sonnet | feat/federation-m3-security-review   | M3-11                             | 10K      | Two review rounds budgeted. PRD requires explicit test for every 401/403 path — review verifies coverage.                                                |
+| FED-M3-13 | not-started | Docs update: `docs/federation/SETUP.md` mTLS handshake section, new `docs/federation/HARNESS.md` for federation-harness usage, OID reference table in SETUP.md, scope enforcement pipeline diagram. Runbook still M7-deferred.                                                                         | #462  | haiku  | feat/federation-m3-docs              | M3-12                             | 5K       | One ASCII diagram for the auth-guard → scope → RBAC pipeline; helps future reviewers reason about denial paths.                                          |
+| FED-M3-14 | not-started | PR aggregate close, CI green, merge to main, close #462. Release tag `fed-v0.3.0-m3`. Update mission manifest M3 row → done; M4 row → in-progress when work begins.                                                                                                                                    | #462  | sonnet | chore/federation-m3-close            | M3-13                             | 3K       | Same close pattern as M1-12 / M2-13.                                                                                                                     |
 
 **M3 estimate:** ~100K tokens (vs MILESTONES.md 40K — same per-task breakdown pattern as M1/M2: tests, review, and docs split out from implementation cost). Largest milestone in the federation mission.
 
@@ -118,6 +118,10 @@ Goal: Two federated gateways exchange real data over mTLS. Inbound requests pass
 
 **Test bed fallback:** If `mos-test-1.woltje.com` / `mos-test-2.woltje.com` are still blocked on `FED-M2-DEPLOY-IMG-FIX` when M3-11 is ready to run, the harness's local `docker-compose.two-gateways.yml` is a sufficient stand-in. Production-host validation moves to M7 acceptance suite (PRD AC-12).
 
+**Backlog sync — 2026-06-24 (orchestrator):** Status reconciled against `origin/main` (release 0.0.48). Landed on main: **FED-M3-01** (DTOs, PR #506), **FED-M3-02** (harness scaffold, PR #505), **FED-M3-03** (mTLS auth-guard, PR #509 — CRIT-1/2 + HIGH-1..4 remediated in-PR), **FED-M3-08** (outbound mTLS client, PR #508). With M3-01/03/08 merged, three cards became dependency-clear and were dispatched to the idle coder lane: **FED-M3-04** scope.service → coder0 (`feat/federation-m3-scope-service`); **FED-M3-09** query-source + **FED-M3-07** capabilities verb → coder1 (`feat/federation-m3-query-source` first). Reviewer warmed for the M3 trust-boundary PRs. Remaining blocked-by-DAG: M3-05/06 (await M3-04), M3-10 (await M3-05/06), M3-11 (await M3-09), M3-12→14 (tail). Deploy chain (DEPLOY-IMG-FIX → 03/04) still independent of M3 code — harness local docker-compose fallback covers M3-11.
+
+**Backlog sync #2 — 2026-06-24 (orchestrator):** **FED-M3-09** (query-source) merged via PR #673 and **FED-M3-07** (capabilities) merged via PR #674 — both squash-merged on independent agent review-of-record + green CI (formal Gitea approve unavailable under the shared service account; merge is not gated by the self-approve guard). **FED-M3-05** (list verb) dispatched to coder1 (based on the M3-04 branch, rebase onto main once #672 lands). **FED-M3-04** (scope.service, PR #672) is in review-changes (one include_personal no-leak test outstanding). **DAG fix:** corrected `FED-M3-11` depends_on from `M3-02, M3-09` → `M3-02, M3-04, M3-05, M3-06, M3-09` — the E2E acceptance cases (#1–#5, #8–#10) exercise list/get over mTLS, so the server verbs + scope service are hard prerequisites; the original edge set omitted them and caused a premature M3-11 dispatch. Note: M3 read-path invariant for M3-11 is **no-persist + existing enrollment audit only** — read-verb audit-log writes are deferred to M4 (see M3-05/06 notes), so M3-11 must not assert read-audit-log entries.
+
 ## Milestone 4 — search + audit + rate limit (FED-M4)
 
 _Deferred. Issue #463._

docs/fleet/north-star.md

@@ -353,6 +353,25 @@ re-evaluate if isolation or write-volume demands it.
 - **Docs as projections:** `docs/TASKS.md` and `MISSION-MANIFEST.md` become generated exports of the DB, not hand-maintained.
 - **Sub-decision pending:** dedicated schema in existing PG instance (recommended) vs. dedicated PG instance. Revisit if isolation or write-volume demands it.
 
+## Decisions of record (2026-06-24, with Jason)
+
+- **Per-agent model switch (operator-configurable, NOT a global lock):** model selection is
+  **per-agent**, never a host-global pin. Claude sessions MUST NOT be locked to a single model in
+  `~/.claude/settings.json`; each agent chooses its model independently. The plumbing already exists —
+  roster `model_hint` → `MOSAIC_AGENT_MODEL` → `start-agent-session.sh` appends `--model <hint>` to that
+  agent's harness (claude or pi); settable today via `mosaic fleet add|edit <agent> --model <hint>`.
+  **North-star target:** surface this as a **per-agent model switch in the webUI** (with CLI/TUI parity
+  per MVP-X1) — read the roster, expose a per-agent model dropdown, write `model_hint` back, and restart
+  that one agent to apply. Unset = inherit the harness default. This **composes with** the budget
+  downgrade ladder (opus → sonnet → haiku, then Claude → Codex): the operator sets the per-agent model
+  _intent/ceiling_; budget pacing may downgrade within policy. Tracked as a Fleet `TASKS.md` entry under
+  the Phase-5 webUI surface.
+- **Orchestrator runtime (confirmed live):** the **orchestrator and enhancer run Claude Opus 4.8 in the
+  Claude Code harness**; only workers (coder/reviewer) run pi/gpt-5.5. Consistent with the 2026-06-20
+  "Claude reserved for Claude Code only" decision (the orchestrator runs _in_ Claude Code, not an
+  alternate Claude harness). Pi/gpt-5.5 as the orchestrator is permitted **only if proven** at least as
+  satisfactory; absent that proof, the orchestrator stays on Claude Opus 4.8.
+
 ## Future enhancements (north-star, post-MVP — not on the MVP track)
 
 - **Mosaic Claude Discord Plugin** — a first-party Mosaic Discord connector that properly

docs/guides/admin-guide.md

@@ -293,13 +293,24 @@ Each OIDC provider requires its client ID, client secret, and issuer URL togethe
 
 ### Plugins
 
-| Variable               | Description                                                                |
-| ---------------------- | -------------------------------------------------------------------------- |
-| `DISCORD_BOT_TOKEN`    | Discord bot token (enables Discord plugin)                                 |
-| `DISCORD_GUILD_ID`     | Discord guild/server ID                                                    |
-| `DISCORD_GATEWAY_URL`  | Gateway URL for Discord plugin to call (default: `http://localhost:14242`) |
-| `TELEGRAM_BOT_TOKEN`   | Telegram bot token (enables Telegram plugin)                               |
-| `TELEGRAM_GATEWAY_URL` | Gateway URL for Telegram plugin to call                                    |
+| Variable                      | Description                                                                                                                               |
+| ----------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
+| `DISCORD_BOT_TOKEN`           | Discord bot token (enables Discord plugin)                                                                                                |
+| `DISCORD_SERVICE_TOKEN`       | Required high-entropy service credential used to authenticate and sign Discord ingress; inject through the approved secret mechanism only |
+| `DISCORD_SERVICE_USER_ID`     | Required Mosaic service-principal user ID that owns persisted Discord conversations; the original Discord user ID remains audit metadata  |
+| `DISCORD_GUILD_ID`            | Discord guild/server ID                                                                                                                   |
+| `DISCORD_GATEWAY_URL`         | Gateway URL for Discord plugin to call (default: `http://localhost:14242`)                                                                |
+| `DISCORD_ALLOWED_GUILD_IDS`   | Required comma-separated Discord guild snowflake allowlist; default-deny                                                                  |
+| `DISCORD_ALLOWED_CHANNEL_IDS` | Required comma-separated Discord channel snowflake allowlist; default-deny                                                                |
+| `DISCORD_ALLOWED_USER_IDS`    | Required comma-separated Discord user snowflake allowlist; default-deny                                                                   |
+| `TELEGRAM_BOT_TOKEN`          | Telegram bot token (enables Telegram plugin)                                                                                              |
+| `TELEGRAM_GATEWAY_URL`        | Gateway URL for Telegram plugin to call                                                                                                   |
+
+### Discord ingress security
+
+When `DISCORD_BOT_TOKEN` is configured, `DISCORD_SERVICE_TOKEN`, `DISCORD_SERVICE_USER_ID`, and all three Discord allowlists are required. Gateway startup fails rather than enabling a broad or unauthenticated remote-control surface. The service user ID identifies a provisioned Mosaic service principal for persistence; the original Discord user ID is retained in ingress audit metadata. The service token is a secret supplied by the approved runtime secret mechanism and is never committed or logged.
+
+Inbound Discord messages must originate from an allowed guild, channel, and user, mention the bot, and carry a signed envelope containing the native Discord message ID and a generated correlation ID. The gateway validates the service identity, envelope signature, and allowlists again before dispatching. Replayed Discord message IDs are rejected during the bounded ingress replay window. Durable inbox/idempotency retention is introduced with Tess durable state.
 
 ### Observability
 

docs/scratchpads/462-fed-m3-04-scope-service.md

@@ -0,0 +1,60 @@
+# Scratchpad — FED-M3-04 Scope Service
+
+## Objective
+
+Implement `apps/gateway/src/federation/server/scope.service.ts` for the M3 inbound federation scope-enforcement pipeline.
+
+## Scope / Constraints
+
+- Task: FED-M3-04, issue #462.
+- Branch: `feat/federation-m3-scope-service` from `origin/main` @ 0.0.48.
+- Pure service: no direct DB access; native RBAC/data access is injected per evaluation call.
+- Reuse `parseFederationScope` from M2-03.
+- Workers do not edit `docs/federation/TASKS.md` per repo AGENTS.md.
+
+## Acceptance Criteria
+
+1. Resource allowlist and `excluded_resources` enforced.
+2. Native RBAC evaluated as `subjectUserId` through an injected evaluator.
+3. Scope filter intersection supports `include_teams` and `include_personal` without widening native RBAC.
+4. `max_rows_per_query` caps requested limits.
+5. Service returns `{ allowed: true, filter }` or a structured deny reason usable by M4 audit.
+6. Unit tests cover every deny path.
+
+## Plan
+
+1. Inspect existing federation scope/schema/auth guard contracts.
+2. Add pure `FederationScopeService` plus typed result/filter/deny interfaces.
+3. Add focused unit tests for happy paths, filter intersection, row cap, and deny paths.
+4. Export/register service for future verb controllers.
+5. Run situational tests, baseline gates, code review, then PR.
+
+## Budget
+
+- Provided model tier: sonnet.
+- Estimate from task row: 10K tokens.
+- Working cap assumption: keep implementation focused to FED-M3-04 surfaces only.
+
+## Progress
+
+- Intake complete; dirty base worktree avoided by creating isolated worktree at `/home/jarvis/src/mosaic-mono-v1-fed-m3-04`.
+- Project PRD and federation task spec reviewed.
+- Added `FederationScopeService` with structured allow/deny result types and injected native RBAC evaluator contract.
+- Added unit coverage for happy path, row cap, filter intersection, and every deny path.
+- Exported/registered the service for upcoming M3 verb controllers.
+
+## Verification Evidence
+
+- `pnpm --filter @mosaicstack/gateway test -- src/federation/server/__tests__/scope.service.spec.ts` — pass (10 tests before review update; 11 tests after adding include_personal no-leak coverage).
+- `pnpm build` — pass (23 successful tasks).
+- `pnpm typecheck` — pass (41 successful tasks; re-run after review update).
+- `pnpm lint` — pass (23 successful tasks; re-run after review update).
+- `pnpm format:check` — pass (re-run after review update).
+- `pnpm test` — pass after starting local `postgres`/`valkey` and running `pnpm --filter @mosaicstack/db db:push` for the DB-backed cross-user isolation suite (41 successful tasks; gateway 477 passed / 11 skipped).
+- Code review: `~/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted` — approve, 0 findings.
+- Security review: `~/.config/mosaic/tools/codex/codex-security-review.sh --uncommitted` — risk none, 0 findings.
+
+## Risks / Blockers
+
+- Issue #462 is already closed in provider output; likely milestone tracking mismatch. Will still reference #462 in PR body unless orchestrator redirects.
+- Local full-test setup required `docker compose up -d postgres valkey` + `db:push`; containers were stopped with `docker compose down` after verification.

docs/scratchpads/561-python-is-python3.md

@@ -0,0 +1,33 @@
+# Issue #561 — Bare python on agent hosts
+
+## Objective
+
+Make the durable bootstrap/provisioning guidance ensure agent hosts provide a bare `python` command that resolves to Python 3.
+
+## Scope
+
+- Add Debian/Ubuntu `python-is-python3` to agent-host prerequisites in bootstrap docs.
+- Check for actual OS package provisioning scripts and update only if an existing agent-host package install path exists.
+- Do not touch live host state.
+- Do not update `docs/TASKS.md`; repo guidance says workers read it but never modify it.
+
+## Recon
+
+- Issue #561 confirms repeated `python: command not found` failures from fleet agents that emit `python foo.py`.
+- `guides/BOOTSTRAP.md` and `packages/mosaic/framework/guides/BOOTSTRAP.md` are the source and packaged framework copies of the bootstrap guide.
+- Targeted repo sweep found no agent-host Debian package provisioning script. Existing `apt-get install` hits are CI/test helper paths or unrelated deployment docs.
+
+## Plan
+
+1. Add a host prerequisite section to both bootstrap guide copies.
+2. Include `python-is-python3` in the Debian/Ubuntu package list with an issue comment.
+3. Note the non-Debian equivalent as a `/usr/bin/python -> python3` symlink.
+4. Validate markdown/diff, run shell syntax checks where applicable, run required review, commit, queue guard, and push.
+
+## Validation Log
+
+- `rg` recon: no existing agent-host Debian package provisioning script; only CI/test helper `apt-get install` paths and unrelated deployment docs.
+- `git diff --check`: passed.
+- `bash -n packages/mosaic/framework/install.sh tools/install.sh packages/mosaic/framework/tools/bootstrap/init-project.sh packages/mosaic/framework/tools/_scripts/mosaic-bootstrap-repo`: passed. No touched shell scripts.
+- `~/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted`: approved, 0 findings.
+- `pnpm format:check`: initially blocked because `node_modules` was absent and `prettier` was unavailable; `pnpm install --frozen-lockfile` initially hit an invalid `/root` pnpm store path. Reran install with `--store-dir /home/hermes/agent-work/.pnpm-store`, then `pnpm format:check` passed.

docs/scratchpads/672-fleet-personas-timeout.md

@@ -0,0 +1,25 @@
+# Scratchpad — fleet-personas spec timeout
+
+## Objective
+
+Raise the `@mosaicstack/mosaic` Vitest timeout to 30s at config level so filesystem-backed fleet drift-guard specs (`fleet-personas`, `fleet-profiles`, and siblings) stop false-reding under contended CI.
+
+## Plan
+
+1. Move timeout policy into `packages/mosaic/vitest.config.ts` with `testTimeout: 30_000`.
+2. Remove the narrower `fleet-personas.spec.ts` local override so PR #677 fixes the suite class, not one file.
+3. Run targeted fleet specs plus typecheck/lint/format gates.
+4. Commit, queue guard, push, PR update.
+
+## Evidence
+
+- `pnpm --filter @mosaicstack/mosaic test -- src/commands/fleet-personas.spec.ts` — pass (8 tests; initial narrow fix).
+- `pnpm typecheck` — pass (41 tasks; initial narrow fix).
+- `pnpm lint` — pass (23 tasks; initial narrow fix).
+- `pnpm format:check` — pass after formatting this scratchpad (initial narrow fix).
+- Package-wide timeout follow-up:
+  - `pnpm --filter @mosaicstack/mosaic test -- src/commands/fleet-personas.spec.ts src/commands/fleet-profiles.spec.ts` — pass (24 tests).
+  - `pnpm --filter @mosaicstack/mosaic test` — pass (44 files / 618 tests).
+  - `pnpm typecheck` — pass (41 tasks).
+  - `pnpm lint` — pass (23 tasks).
+  - `pnpm format:check` — pass.

docs/scratchpads/703-wrapper-interactive-auth.md

@@ -0,0 +1,49 @@
+# #703 Git Wrapper Interactive and Auth Resilience
+
+## Objective
+
+Restore the deployed Git wrapper contract: issue-create supports interactive invocation and Gitea mutation behavior tolerates a stale Tea authenticated user by validating current identity and using the existing host-scoped API fallback.
+
+## Scope
+
+- `packages/mosaic/framework/tools/git/issue-create.sh`
+- `packages/mosaic/framework/tools/git/detect-platform.sh`
+- Git wrapper regression harnesses
+- This scratchpad
+
+## Requirements / acceptance evidence
+
+1. `issue-create -i` and `--interactive` prompt for missing issue fields without exposing credentials.
+2. Explicit command-line fields retain precedence and do not trigger prompt input.
+3. Gitea wrapper resolves the current user dynamically from the target host and does not rely on the saved Tea user identity.
+4. A Tea `GetUserByName` failure falls back to authenticated API creation.
+5. Existing body-safety, login-resolution, issue-create, and pr-create paths remain green.
+6. Source framework is re-seeded to deployed `~/.config/mosaic`, then deployed wrappers are verified end to end.
+
+## Plan
+
+1. Add failing shell regression harness for interactive input and stale Tea user fallback.
+2. Implement minimal helper and parser changes.
+3. Run wrapper harnesses, syntax checks, and repository baseline checks.
+4. Re-seed deployed framework and run live wrapper verification.
+5. Commit, queue guard, push, open PR, and stop for independent review.
+
+## Progress
+
+- Issue #703 filed before code; issue comment records #536 root cause and stale-login trigger.
+- Deployed wrapper `issue-create.sh -i` reproduced: `Unknown option: -i` (exit 1).
+- Live Tea mutation did not reproduce `GetUserByName` on this host because the current mosaicstack Tea login is valid. The test harness models the reported stale authenticated-user condition.
+- Implemented `-i` / `--interactive` prompt collection and a dynamic Tea `/user` validation. A stale Tea identity now selects the existing host-scoped Gitea API fallback before mutation for both issue and PR creation.
+- Re-seeded the framework with `MOSAIC_INSTALL_MODE=keep MOSAIC_SYNC_ONLY=1 bash packages/mosaic/framework/install.sh`. Installed and source wrapper SHA-256 values matched.
+- Live deployed verification: interactive issue-create opened then closed #704; installed dynamic identity resolved `jason.woltje`.
+
+## Verification
+
+- PASS: `packages/mosaic/framework/tools/git/test-issue-create-interactive-auth.sh`
+- PASS: `packages/mosaic/framework/tools/git/test-issue-create-body-safety.sh`
+- PASS: `packages/mosaic/framework/tools/git/test-gitea-login-resolution.sh`
+- PASS: `packages/mosaic/framework/tools/git/test-pr-metadata-gitea.sh`
+- PASS: `packages/mosaic/framework/tools/git/test-pr-merge-gitea-empty-uid.sh`
+- PASS: `bash packages/mosaic/framework/tools/git/test-lane-brief-pr-linkage.sh`
+- PASS: `bash -n packages/mosaic/framework/tools/git/*.sh`
+- PASS: Prettier check for this scratchpad

docs/scratchpads/FED-M3-05-list-verb.md

@@ -0,0 +1,52 @@
+# FED-M3-05 — Federation List Verb Scratchpad
+
+## Objective
+
+Implement `POST /api/federation/v1/list/:resource`.
+
+## Scope
+
+- Wire `FederationAuthGuard` → `FederationScopeService` → read-only list query layer.
+- Apply `max_rows_per_query` row cap and return pagination metadata when truncated.
+- Tag returned rows with `_source: "local"`.
+- Keep audit writes deferred to M4.
+- No request/response body persistence.
+
+## Base / branch
+
+- Branch: `feat/federation-m3-verb-list`
+- Base: `main` after M3-04 scope service merged via PR #672 (`c739256a`).
+
+## Implementation notes
+
+- Added `ListController` under `apps/gateway/src/federation/server/verbs/`.
+- Added `FederationListQueryService` as the read-only query layer and native RBAC evaluator.
+- Query resources supported in M3 list path:
+  - `tasks`: project/mission scoped tasks visible through personal/team project access.
+  - `notes`: non-empty `mission_tasks.notes` rows visible through personal/team mission access.
+  - `memory`: user-owned `insights` and `preferences` rows.
+  - `credentials` / `api_keys`: denied by native RBAC in M3 even if present in scope; sensitive-resource implementation is not part of FED-M3-05.
+- Cursor pagination uses an opaque base64url keyset cursor over `(createdAt, id)`; DB reads fetch at most `limit + 1` rows per resource query.
+- Reviewer isolation fix: `mission_tasks.notes` rows are always constrained by `missionTasks.userId = subjectUserId` and accessible mission IDs; team scope narrows missions but never widens to other users' mission task notes.
+- Follow-up review fix: memory listing now uses deterministic table-block pagination (`insights` first, then `preferences`) with cursor source metadata, so one table's cursor is never applied to the other.
+- Follow-up hardening: missing auth-guard context returns a structured federation `unauthorized` envelope; unsupported resources and non-encodable truncated cursors throw instead of silently crashing/truncating.
+
+## Tests
+
+- `pnpm --filter @mosaicstack/gateway test -- list.controller.spec.ts list-query.service.spec.ts` — PASS (16 tests, including PGlite regression coverage for team-scoped notes isolation, unauthorized mission notes exclusion, `includePersonal: false`, deterministic memory pagination, missing context envelope, unsupported resource, and cursor encode failure).
+- `pnpm --filter @mosaicstack/gateway typecheck` — PASS.
+- `pnpm --filter @mosaicstack/gateway lint` — PASS.
+- `pnpm format:check` — PASS.
+- `pnpm typecheck` — PASS (41/41 turbo tasks).
+- `pnpm lint` — PASS (23/23 turbo tasks).
+- `pnpm --filter @mosaicstack/gateway test` — FAIL in pre-existing/live-DB integration suite: `apps/gateway/src/__tests__/cross-user-isolation.test.ts` cleanup cannot connect to local PostgreSQL on `localhost:5433`. New list tests pass; failure is outside FED-M3-05.
+
+## Review evidence
+
+- `~/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted` — PASS after follow-up remediation; approve, no findings.
+- `~/.config/mosaic/tools/codex/codex-security-review.sh --uncommitted` — PASS after follow-up remediation; risk level none, no findings.
+- Security-review note: read-path audit logging remains intentionally deferred to M4 per orchestrator clarification and FED-M3-05 scope.
+
+## Risks / follow-up
+
+- Read-path audit logging remains intentionally deferred to M4.

docs/scratchpads/FED-M3-07-capabilities.md

@@ -0,0 +1,65 @@
+# FED-M3-07 — Capabilities Verb Scratchpad
+
+## Objective
+
+Implement `GET /api/federation/v1/capabilities` in `apps/gateway/src/federation/server/verbs/capabilities.controller.ts`.
+
+## Scope
+
+- Add read-only capabilities controller under federation server verbs.
+- Use `FederationAuthGuard` only; active grant is sufficient and no native RBAC/scope-service eval runs.
+- Response shape: `{ resources, excluded_resources, max_rows_per_query, supported_verbs }` derived from grant scope.
+- Register controller in `FederationModule`.
+- Unit-test happy path, defaults, no-context guard seam, and invalid scope handling.
+
+## Constraints / assumptions
+
+- Issue: #462.
+- Branch: `feat/federation-m3-verb-capabilities` from `origin/main` (`3eeed04e`).
+- Depends on M3-03 auth guard; guard attaches `request.federationContext.scope` after active-grant validation.
+- ASSUMPTION: `supported_verbs` is the M3 verb set from `@mosaicstack/types` (`list`, `get`, `capabilities`).
+- ASSUMPTION: `filters`/`rate_limit` are intentionally omitted for FED-M3-07 because the card’s response shape lists only the four required fields.
+- Budget: no explicit hard cap from orchestrator; working cap ~4K-8K tokens for card implementation + tests + PR cycle.
+
+## Plan
+
+1. Write controller unit tests first.
+2. Implement controller and module registration.
+3. Run scoped tests + typecheck/lint/format.
+4. Run Codex code/security review and remediate.
+5. Commit, queue guard, push, PR via wrapper.
+
+## Progress
+
+- 2026-06-24: Intake complete; fresh worktree created from origin/main.
+- 2026-06-24: Added `CapabilitiesController`, registered it in `FederationModule`, and added 5 unit tests.
+- 2026-06-24: Code/security reviews passed with no findings.
+
+## Tests run
+
+- `pnpm --filter @mosaicstack/gateway test -- capabilities.controller.spec.ts` — PASS (5 tests).
+- `pnpm --filter @mosaicstack/gateway typecheck` — PASS.
+- `pnpm --filter @mosaicstack/gateway lint` — PASS.
+- `pnpm format:check` — PASS.
+- `pnpm typecheck` — PASS (41/41 turbo tasks).
+- `pnpm lint` — PASS (23/23 turbo tasks).
+- `pnpm test` — FAIL in pre-existing/live-DB integration suite: `apps/gateway/src/__tests__/cross-user-isolation.test.ts` cleanup hit PostgreSQL connection/schema state for the `messages` table. Changed capabilities tests passed; failure is outside FED-M3-07 surface. No `fleet-personas.spec` flake encountered.
+
+## Review evidence
+
+- `~/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted` — PASS/approve, no findings.
+- `~/.config/mosaic/tools/codex/codex-security-review.sh --uncommitted` — PASS, risk level none, no findings.
+
+## Risks / blockers
+
+- Full repo `pnpm test` may hit known `fleet-personas.spec` flake per orchestrator; ignore that specific flake if encountered.
+- Previous card saw local DB schema issue in `cross-user-isolation.test.ts`; scoped capabilities tests should be authoritative for this surface.
+
+## Acceptance evidence mapping
+
+| Acceptance criterion                                                       | Evidence                                                                                                                   |
+| -------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------- |
+| GET `/api/federation/v1/capabilities` exists                               | Route metadata test in `capabilities.controller.spec.ts`; scoped test PASS                                                 |
+| Uses active-grant auth guard and no RBAC eval                              | Guard metadata test confirms only `FederationAuthGuard`; controller has no service injections/RBAC calls; scoped test PASS |
+| Response enumerates resources/excluded/max rows/supported verbs from scope | Happy-path/default scope tests + response schema parse; scoped test PASS                                                   |
+| Read-only/no persistence side effects                                      | Controller only parses request `federationContext.scope` and returns a DTO; no DB/service dependency; code review PASS     |

docs/scratchpads/FED-M3-09-query-source.md

@@ -0,0 +1,67 @@
+# FED-M3-09 — Query Source Service Scratchpad
+
+## Objective
+
+Implement `apps/gateway/src/federation/client/query-source.service.ts` for `source: "local" | "federated:<host>" | "all"` routing.
+
+## Scope
+
+- Add QuerySourceService in gateway federation client layer.
+- Unit-test local-only, single federated peer, all-source fan-out/merge, and per-peer partial failures.
+- Keep `docs/federation/TASKS.md` read-only per project agent guidance.
+
+## Constraints / assumptions
+
+- Issue: #462.
+- Branch: `feat/federation-m3-query-source` from `origin/main` (`e0e7be70`).
+- ASSUMPTION: `federated:<host>` should match active outbound peers by `commonName` first and by `endpointUrl` host/hostname as compatibility fallback; source tags use `peer.commonName` per `@mosaicstack/types` source-tag docs.
+- ASSUMPTION: QuerySourceService provides list/fan-out behavior; get/source routing can be layered later because card acceptance says merge rows.
+- ASSUMPTION: `source: "all"` cannot safely return a single continuation cursor for multiple sub-sources; any subquery cursor marks the merged response `_partial: true` + `_truncated: true` while omitting `nextCursor`.
+- Budget: no explicit hard cap from orchestrator; working cap ~8K-12K tokens for card 1 implementation + tests + PR cycle.
+- OpenBrain unavailable: credential loader failed with missing `/home/jarvis/.config/mosaic/credentials.json`; not blocking code delivery.
+
+## Plan
+
+1. Review federation client/types/db patterns.
+2. Write unit tests for source behavior.
+3. Implement QuerySourceService and export/register it in FederationModule.
+4. Run scoped tests, typecheck, lint, format.
+5. Run codex uncommitted review and remediate.
+6. Commit, queue guard, push, PR via wrapper.
+
+## Progress
+
+- 2026-06-24: Intake complete; using isolated worktree to avoid dirty orchestrator files in original checkout.
+- 2026-06-24: Added QuerySourceService, module export, barrel export, and 7 unit tests.
+- 2026-06-24: First Codex review found pagination and port-host matching issues; both remediated with tests.
+
+## Tests run
+
+- `pnpm --filter @mosaicstack/gateway test -- query-source.service.spec.ts` — PASS (7 tests).
+- `pnpm --filter @mosaicstack/gateway typecheck` — PASS.
+- `pnpm --filter @mosaicstack/gateway lint` — PASS.
+- `pnpm format:check` — PASS.
+- `pnpm typecheck` — PASS (41/41 turbo tasks).
+- `pnpm lint` — PASS (23/23 turbo tasks).
+- `pnpm test` — FAIL in pre-existing/live-DB integration suite: `apps/gateway/src/__tests__/cross-user-isolation.test.ts` cleanup hit `relation "messages" does not exist` against local PostgreSQL. Changed QuerySource unit tests passed; failure is outside FED-M3-09 surface and appears tied to local DB schema state.
+
+## Review evidence
+
+- `~/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted` — first pass request-changes, 2 should-fix findings (all-source cursor handling; endpoint port host matching).
+- Remediation: `_partial` + `_truncated` when any all-source subquery has `nextCursor`; endpoint match accepts URL `host` and `hostname`; added tests for both.
+- `~/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted` — PASS/approve, no findings.
+- `~/.config/mosaic/tools/codex/codex-security-review.sh --uncommitted` — PASS, risk level none, no findings.
+
+## Risks / blockers
+
+- Federation query layer is not yet wired; service API needs to be stable and easy to compose.
+- Must avoid hard-failing `source: all` on remote peer failures.
+
+## Acceptance evidence mapping
+
+| Acceptance criterion                                                            | Evidence                                                                          |
+| ------------------------------------------------------------------------------- | --------------------------------------------------------------------------------- |
+| local source returns local rows tagged `_source: local`                         | `query-source.service.spec.ts` local test; scoped test PASS                       |
+| `federated:<host>` queries selected peer and tags rows with peer source         | `query-source.service.spec.ts` commonName/endpoint-host tests; scoped test PASS   |
+| `all` fans out local + active outbound peers in parallel and merges tagged rows | `query-source.service.spec.ts` all-source call-order/merge test; scoped test PASS |
+| per-peer failure on `all` returns `_partial: true`, not throw                   | `query-source.service.spec.ts` peer failure test; scoped test PASS                |

docs/scratchpads/tess-20260712.md

@@ -0,0 +1,58 @@
+# Scratchpad — Tess Interaction Agent
+
+## 2026-07-12 — Mission intake
+
+**Objective:** Build a Pi-native GPT-5.6 Sol high-reasoning Mosaic interaction agent, named Tess, as Jason's primary Discord/CLI access point for Mosaic fleet and transitional Hermes capabilities. Tess complements Mos and must not become a competing orchestrator.
+
+**Issue:** #706
+
+**Budget:** No explicit cap provided. Original working estimate was 290K implementation/review tokens. That estimate is superseded after six security prerequisite tasks were added; revised arithmetic total is pending because the calculation tool was blocked by runtime consent. Run at most two workers; prefer one implementation lane plus one independent review/discovery lane. Re-estimate after planning approval and each milestone.
+
+**Evidence gathered:**
+- Mosaic already provides Pi lifecycle hooks, fleet/tmux sessions, Matrix connector/controller pieces, typed chat events, Discord/Telegram channel plugins, and command/plugin registries.
+- Current `IProviderAdapter` is an LLM model/completion abstraction, not an external agent/session provider.
+- Required new seam is `AgentRuntimeProvider`: sessions, stream, message, terminate, hierarchy, attach, health, capabilities.
+- Recurring cross-runtime needs: unified memory/retrieval, Discord routing/approvals, agent state/inbox/compaction recovery, runtime bootstrap, fleet/incident controls, and GitOps workflow.
+- Project truth must remain in canonical project/Mosaic stores; semantic memory is retrieval/mirror.
+
+**Decisions:**
+1. Name: Tess (tessera). Stable machine key `tess`; display name configurable.
+2. Mos owns orchestration; Tess delegates Mos-owned work through an explicit coordination contract.
+3. Gateway owns ingress/auth/routing; Discord and CLI remain thin clients.
+4. tmux/fleet ships first behind an adapter; Matrix/native Mosaic is the forward transport.
+5. Hermes integration is transitional and capability-negotiated; unsupported operations fail closed.
+6. No unrestricted Discord shell. Privileged/destructive/customer-visible actions require authorization and approval.
+
+**Plan:**
+1. Land requirements/architecture/task graph.
+2. Deliver runtime contracts and security model.
+3. Deliver durable Pi service/state.
+4. Deliver Discord and CLI.
+5. Deliver fleet/Mos/Hermes/memory/tool plugins.
+6. Deliver Matrix/native transport, migration matrix, recovery, docs, and qualification.
+
+**Progress:** Issue #706 created. PRD/manifest/tasks initialized on clean branch `feat/tess-interaction-agent` from `origin/main`.
+
+**Risks:** 14 GB root filesystem headroom; active fleet lanes; broad migration scope; Discord privilege boundary; possible duplicate orchestration authority.
+
+## 2026-07-12 — Independent planning and threat review
+
+**Verdict received:** BLOCK TESS-PLAN-001. Coding remains stopped.
+
+**Blocking findings:** formal threat model absent; verification matrix absent; migration inventory implied but absent; non-existent task paths; AC-TESS-03 lacked a crisp test. Security review also identified command scope bypass, cross-tenant session attachment, MCP actor impersonation, unsafe Discord service ingress, pre-persistence/egress secret leakage, non-durable replay, and globally scoped GC.
+
+**Remediation applied:**
+- Added `docs/tess/ARCHITECTURE.md`.
+- Added `docs/tess/THREAT-MODEL.md` with TM-01..12.
+- Added `docs/tess/VERIFICATION-MATRIX.md` mapping AC-TESS-01..11.
+- Added `docs/tess/MIGRATION-INVENTORY.md`.
+- Added hard requirements TESS-SEC-002..009.
+- Added six prerequisite security tasks before provider/ingress implementation.
+- Corrected task paths to existing package surfaces.
+- Added explicit GPT-5.6 Sol/high/tool-policy status verification for AC-TESS-03.
+
+**Re-review 1:** BLOCK only on composite `repo` values that looked like nonexistent paths. Remediated by declaring comma-separated roots and validating every root.
+
+**Final focused review:** PASS. Deterministic audit validated all task repository roots with zero missing paths; no planning placeholders remained; security prerequisites still gate Tess exposure; observability traceability is explicit.
+
+**Current gate:** planning PR must merge to `main` with terminal-green CI before any source-code worker starts.

docs/scratchpads/tess-m1-sec-004-discord-ingress.md

@@ -0,0 +1,35 @@
+# Scratchpad — TESS-M1-SEC-004 Discord ingress
+
+- **Task / issue:** TESS-M1-SEC-004 / #707
+- **Branch:** `fix/tess-discord-ingress` from `origin/main` at `59e49cfd`
+- **Objective:** Authenticate the Discord plugin service at gateway ingress; enforce explicit guild/channel/user allowlists; attach Discord message and generated correlation IDs; reject replayed native message IDs.
+- **Scope:** `plugins/discord`, `apps/gateway`, and existing Discord admin/developer protocol docs.
+- **Budget:** Task estimate 28K; no explicit hard cap supplied.
+- **Assumptions:** The Discord plugin and gateway share an injected high-entropy `DISCORD_SERVICE_TOKEN`; a configured Discord plugin fails closed without it. Allowlist configuration is comma-separated Discord snowflakes. Discord native message ID is the replay key, with bounded in-memory retention pending the M2 durable inbox/idempotency work.
+
+## Plan
+
+1. Add failing tests covering ingress service authentication/signing, unlisted guild/channel/user rejection, correlation propagation, and replay rejection.
+2. Implement the signed Discord ingress envelope and allowlist validation in the plugin.
+3. Authenticate and validate the envelope at the gateway boundary, then enforce bounded replay protection before agent dispatch.
+4. Document the service-token and allowlist operations; run focused and baseline gates; obtain independent review.
+
+## Progress
+
+- 2026-07-12: Intake complete; PRD TESS-SEC-005, architecture, and threat model reviewed.
+- Added service-token Socket.IO authentication, HMAC-signed Discord envelopes, default-deny guild/channel/user allowlists, correlated message metadata, bounded replay rejection, and fail-fast configuration checks.
+- Code and security reviews completed. Code review findings on service persistence ownership, package-boundary tests, disconnected ingress observability, and chat payload validation were remediated; final independent code review approved.
+
+## Risks / blockers
+
+- Existing `main` has known unrelated Prettier debt; only changed files will be held format-clean. Durable replay persistence is intentionally out of scope for this M1 prerequisite and belongs to TESS-M2 durable inbox/idempotency work.
+
+## Verification evidence
+
+- Focused ingress suite: `pnpm --filter @mosaicstack/gateway test -- discord-ingress.security.spec.ts` — 7 passed.
+- Gateway suite: `pnpm --filter @mosaicstack/gateway test` — 513 passed, 11 skipped.
+- Plugin suite: `pnpm --filter @mosaicstack/discord-plugin test` — no tests, passed by configured `--passWithNoTests`.
+- `pnpm typecheck` — passed.
+- `pnpm lint` — passed.
+- `pnpm format:check` — fails only on the known pre-existing Tess documentation debt listed in the task dispatch; changed files pass targeted Prettier verification.
+- Codex security review — no findings; final Codex code review — approved.

docs/tess/ARCHITECTURE.md

@@ -0,0 +1,71 @@
+# Tess Architecture
+
+## Purpose
+
+Tess is the Mosaic operator interaction plane. Mos remains the coding/general fleet orchestration authority. Tess receives authorized operator intent, presents fleet/session state, delegates Mos-owned work to Mos, and exposes native Mosaic plus transitional external-agent capabilities through normalized providers.
+
+## Component Boundaries
+
+```text
+Discord plugin ─┐
+                ├─ authenticated ingress envelope ─> Mosaic Gateway
+mosaic tess CLI ┘                                  │
+                                                   ├─ policy/approval/audit
+                                                   ├─ Tess durable session service (Pi GPT-5.6 Sol high)
+                                                   ├─ AgentRuntimeProvider registry
+                                                   │   ├─ native Pi provider
+                                                   │   ├─ fleet/tmux provider
+                                                   │   ├─ Hermes adapter
+                                                   │   └─ Matrix/native transport provider
+                                                   ├─ memory/state/inbox plugins
+                                                   └─ Mos coordination adapter ─> Mos / fleet queue
+```
+
+## Core Contract
+
+`AgentRuntimeProvider` is separate from the existing model-completion `IProviderAdapter`. It normalizes external and native agent runtimes without leaking provider-specific schemas.
+
+Required operations:
+
+- `capabilities()` and `health()`
+- `listSessions(scope)`
+- `getSessionTree(scope)`
+- `streamSession(sessionRef, cursor, scope)`
+- `sendMessage(sessionRef, message, idempotencyKey, scope)`
+- `attach(sessionRef, mode, scope)` / `detach()`
+- `terminate(sessionRef, approvalRef, scope)`
+
+Every call receives an immutable, server-derived actor/tenant/channel scope and correlation ID. Caller-supplied actor IDs are forbidden. Unsupported capabilities fail closed with typed errors.
+
+## Authority Model
+
+| Intent                                                                      | Owner                   | Tess behavior                                                          |
+| --------------------------------------------------------------------------- | ----------------------- | ---------------------------------------------------------------------- |
+| Conversation, status, retrieval, safe diagnostics                           | Tess                    | Execute within policy                                                  |
+| Code/project decomposition, worker assignment, reviews, merge orchestration | Mos                     | Create a correlated handoff and observe result                         |
+| Destructive, privileged, external/customer-visible action                   | Human approval + policy | Propose, wait for durable one-time approval, then execute idempotently |
+| Provider-specific unsupported action                                        | None                    | Fail closed; never emulate silently                                    |
+
+## Session and State Model
+
+A Tess session has stable `sessionId`, `tenantId`, `ownerId`, provider/runtime identity, ingress bindings, cursor, checkpoint, inbox/outbox, and idempotency records. Discord and CLI bind to the same authorized session. Ownership is verified server-side on every list/read/attach/send/terminate operation.
+
+Valkey may hold ephemeral coordination state; PostgreSQL is canonical for durable session bindings, approvals, audit, checkpoints, inbox/outbox, and idempotency. Pi session files are replay sources, not cross-agent truth.
+
+## Transport Strategy
+
+- **Initial:** fleet/tmux provider, including exact target, socket, identity, heartbeat, and safe attach semantics.
+- **Forward:** Matrix/native Mosaic provider using authenticated identity, idempotent transaction IDs, replay cursors, and the same contract suite.
+- Discord/CLI never call tmux or Matrix directly.
+
+## Plugin Families
+
+1. Channel: Discord now; other channels later.
+2. Runtime: Pi, fleet/tmux, Hermes, Matrix/native.
+3. Operator tools: fleet health, Mos handoff, GitOps wrappers, incident-safe diagnostics.
+4. Memory/state: search/recent/capture, durable inbox, checkpoint, handoff, compaction recovery.
+5. Migration: capability inventory, adapters, cutover, rollback, telemetry.
+
+## Deployment
+
+Tess runs as a rostered, systemd-supervised Pi agent using GPT-5.6 Sol and high reasoning. Secrets are supplied through approved runtime secret mechanisms. Startup fails when required model, gateway identity, Discord binding, or durable-state dependencies are missing. Health reports effective model/reasoning/tool policy without credential material.

docs/tess/MIGRATION-INVENTORY.md

@@ -0,0 +1,34 @@
+# Tess Capability Migration Inventory
+
+Status values: `native` · `adapt` · `defer` · `reject`. This is the initial inventory; M5 requires implementation and evidence fields to be completed before cutover.
+
+| Capability                               | Current source                           | Target                                     | Initial status | Cutover/rollback intent                                                 |
+| ---------------------------------------- | ---------------------------------------- | ------------------------------------------ | -------------- | ----------------------------------------------------------------------- |
+| Interactive agent chat/session streaming | Hermes/Pi/OpenClaw                       | Mosaic Tess session service                | native         | Dual-run per channel; revert binding to legacy gateway                  |
+| Discord dedicated-channel routing        | Hermes/Claude/OpenClaw plugins           | Mosaic Discord plugin + gateway            | native         | Per-channel binding switch; legacy bot disabled only after soak         |
+| CLI/TUI session interaction and attach   | Hermes/Pi/tmux                           | `mosaic tess` + AgentRuntimeProvider       | native         | Keep direct tmux attach as break-glass rollback                         |
+| Session list/tree/send/terminate         | Hermes/fleet                             | AgentRuntimeProvider                       | native         | Capability-negotiated adapter remains during migration                  |
+| Mos/fleet orchestration handoff          | tmux messaging/Mosaic fleet              | Mosaic coord/fleet provider                | native         | tmux handoff remains initial transport                                  |
+| Kanban/projects/tasks                    | Hermes Kanban                            | Mosaic queue/coord/project providers       | adapt          | Read projection first; mutating cutover after parity/audit              |
+| Skills catalog/load/manage               | Hermes skills/Pi skills                  | Mosaic skill registry/provider             | adapt          | Import metadata/provenance; preserve source skill until validated       |
+| Tools and MCP                            | Hermes/OpenClaw/MCP                      | Mosaic tool registry/MCP                   | adapt          | Default deny; migrate allowlisted tools one capability at a time        |
+| Cron/scheduled work                      | Hermes cron                              | Mosaic scheduler/queue                     | adapt          | Shadow schedules; prevent duplicate execution; rollback owner field     |
+| Memory search/recent/capture             | jarvis-brain/OpenViking/OpenBrain/Hermes | Mosaic memory provider                     | adapt          | Flat/project stores remain truth; semantic systems are mirrors          |
+| User/profile preferences                 | Hermes memory/user profile               | Mosaic user/memory domain                  | adapt          | Provenance + explicit conflict rules; exportable rollback snapshot      |
+| Agent state/inbox/handoff                | OpenClaw extensions/session files        | Mosaic durable state service               | native         | Read legacy handoff during coexistence; write Mosaic only after cutover |
+| Runtime contract/bootstrap               | Mosaic framework/Hermes/OpenClaw         | Mosaic compose/runtime provider            | native         | Legacy launchers remain until clean-host parity passes                  |
+| Repository/PR workflow                   | Mosaic wrappers/Hermes tools             | Mosaic operator plugin                     | native         | Wrapper-only; no raw-provider fallback                                  |
+| Incident-safe diagnostics                | Hermes skills/tools                      | Mosaic scoped operator plugin              | adapt          | Read-only first; privileged recovery requires approval                  |
+| Broad unrestricted shell from Discord    | Hermes/OpenClaw configurations           | None                                       | reject         | No cutover; replace with allowlisted typed operations                   |
+| Raw full transcript bulk migration       | Hermes/Claude/OpenClaw histories         | Indexed summaries/selective import         | reject         | Keep source archives subject to retention; no automatic copy            |
+| Voice/video interaction                  | Hermes optional tools                    | Future Mosaic channel plugins              | defer          | Not required for Tess operational release                               |
+| Matrix transport                         | Mosaic connector                         | AgentRuntimeProvider Matrix implementation | native         | Non-default until contract/reliability parity; tmux rollback            |
+
+## Cutover Gates
+
+1. Capability contract and security tests pass.
+2. Data mapping/provenance and retention are documented.
+3. Shadow or dual-run shows no unauthorized access, loss, or duplicate effects.
+4. Operator runbook and rollback are exercised.
+5. Channel/provider binding changes are reversible without schema rollback.
+6. Legacy capability is disabled only after a defined soak period and evidence review.

docs/tess/MISSION-MANIFEST.md

@@ -0,0 +1,46 @@
+# Mission Manifest — Tess Interaction Agent
+
+## Mission
+
+- **ID:** tess-20260712
+- **Issue:** #706
+- **Branch:** `feat/tess-interaction-agent`
+- **Phase:** Execution
+- **Current Milestone:** TESS-M1 — Runtime contracts and security foundation
+- **Progress:** 0 / 5 delivery milestones complete
+- **Status:** active
+- **Owner:** Mosaic orchestrator; Mos is coordinating fleet authority
+- **Source PRD:** `docs/PRD.md` — `TESS-*` requirements
+- **Scratchpad:** `docs/scratchpads/tess-20260712.md`
+
+## Mission Statement
+
+Ship Tess as Jason's durable Pi-native GPT-5.6 Sol high-reasoning interaction agent for Discord and CLI, with safe visibility/control of Mosaic fleet and transitional Hermes capabilities, while Mos remains the coding/general orchestration authority.
+
+## Invariants
+
+1. Mosaic is the enterprise AI hub; Hermes is a reference migration adapter.
+2. Gateway is the single API surface.
+3. Mos owns coding/general fleet orchestration; Tess owns human interaction, visibility, mediation, and migration access.
+4. Runtime, transport, channel, memory, and external-agent integrations are replaceable adapters.
+5. No source task completes before merged PR, terminal-green CI, independent review, and linked task/issue closure.
+
+## Milestones
+
+| ID      | Issue | Name                                                                | Status      | Exit gate                                                                                                         |
+| ------- | ----- | ------------------------------------------------------------------- | ----------- | ----------------------------------------------------------------------------------------------------------------- |
+| TESS-M1 | #707  | Runtime contracts and security foundation                           | ready       | AgentRuntimeProvider, normalized events/capabilities/errors, RBAC/audit contracts and contract tests merged       |
+| TESS-M2 | #708  | Durable Pi Tess service and state                                   | not-started | GPT-5.6 Sol high service starts, resumes, checkpoints, and passes restart/compaction tests                        |
+| TESS-M3 | #709  | Discord and CLI interaction surfaces                                | not-started | One durable session works through dedicated Discord binding and `mosaic tess`, including attach and approvals     |
+| TESS-M4 | #710  | Fleet, Mos, Hermes, memory, state, and tool plugins                 | not-started | Fleet/Mos boundary and transitional capability matrix demonstrated end-to-end                                     |
+| TESS-M5 | #711  | Matrix/native migration, recovery, documentation, and qualification | not-started | Transport parity, migration/rollback matrix, security review, docs, greenfield and deployment validation complete |
+
+## Success Criteria
+
+All `AC-TESS-*` criteria in `docs/PRD.md` are mapped to reproducible evidence. The final operational test must prove Discord + CLI session continuity, fleet/Mos coordination, authorized Hermes transition capabilities, denial/audit paths, restart recovery, and rollback.
+
+## Session History
+
+| Session | Date       | Runtime              | Outcome                                                                                                                                     |
+| ------- | ---------- | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------- |
+| S1      | 2026-07-12 | Hermes / GPT-5.6 Sol | User commission captured; Mosaic/OpenViking/session/code archaeology completed; issue #706 created; PRD and task control plane initialized. |

docs/tess/TASKS.md

@@ -0,0 +1,34 @@
+# Tasks — Tess Interaction Agent
+
+> Mission: `tess-20260712` · Issue: #706 · PRD requirements: `TESS-*`
+> Orchestrator is sole writer. Workers must not modify this file.
+> `repo` contains one or more comma-separated repository-relative roots; every listed root must exist before dispatch.
+
+| id | status | description | issue | agent | repo | branch | depends_on | estimate | notes |
+| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
+| TESS-PLAN-001 | done | Finalize PRD, architecture, authority boundary, threat model, migration inventory, and verification matrix | #706 | sonnet | docs, packages/types, apps/gateway | feat/tess-interaction-agent | — | 22K | Independent gate PASS after two remediation rounds; completion effective when planning PR merges |
+| TESS-M1-SEC-001 | not-started | Enforce command scopes/roles and durable exact-action approval for privileged/destructive commands | #707 | codex | apps/gateway | fix/tess-command-authz | TESS-PLAN-001 | 25K | TESS-SEC-002; security TDD |
+| TESS-M1-SEC-002 | not-started | Enforce owner/tenant binding on session list/read/attach/send/terminate across REST and WS | #707 | codex | apps/gateway | fix/tess-session-ownership | TESS-PLAN-001 | 30K | TESS-SEC-003; security TDD |
+| TESS-M1-SEC-003 | not-started | Bind MCP actor/tenant to authenticated context and add per-tool scopes | #707 | codex | apps/gateway | fix/tess-mcp-identity | TESS-PLAN-001 | 22K | TESS-SEC-004; security TDD |
+| TESS-M1-SEC-004 | not-started | Add authenticated Discord service ingress, allowlists, correlation and replay protection | #707 | codex | plugins/discord, apps/gateway | fix/tess-discord-ingress | TESS-PLAN-001 | 28K | TESS-SEC-005; security TDD |
+| TESS-M1-SEC-005 | not-started | Redact/classify secret and PII before persistence/egress; harden provider login flow | #707 | codex | apps/gateway, packages/log | fix/tess-redaction | TESS-PLAN-001 | 28K | TESS-SEC-006; seeded canary tests |
+| TESS-M1-SEC-006 | not-started | Scope session GC/retention or separate authorized global retention job | #707 | codex | apps/gateway, packages/log | fix/tess-session-gc-scope | TESS-PLAN-001 | 18K | TESS-SEC-009; isolation TDD |
+| TESS-M1-001 | not-started | Define AgentRuntimeProvider, capabilities, session tree, normalized stream events/errors, attach semantics | #707 | codex | packages/types, packages/agent | feat/tess-runtime-contract | TESS-PLAN-001 | 25K | TESS-ARP-001, TESS-TRN-001; contract TDD |
+| TESS-M1-002 | not-started | Implement provider registry/service with immutable actor scope, approval, audit and correlation boundaries | #707 | codex | apps/gateway, packages/agent | feat/tess-provider-registry | TESS-M1-001,TESS-M1-SEC-001,TESS-M1-SEC-002,TESS-M1-SEC-003 | 30K | TESS-SEC-001..004,007; security TDD |
+| TESS-M1-003 | not-started | Implement tmux/fleet runtime provider and safe attach/message/terminate capability policy | #707 | codex | packages/mosaic, packages/agent | feat/tess-fleet-provider | TESS-M1-002 | 30K | TESS-FLT-001; exact target/identity tests |
+| TESS-M1-OBS-001 | not-started | Implement correlation propagation, structured runtime/provider/tool audit, health/readiness and safe effective-policy status | #707 | codex | apps/gateway, packages/agent, packages/log | feat/tess-observability | TESS-M1-002 | 24K | TESS-OBS-001; no credential material |
+| TESS-M1-V | not-started | Independent architecture/security review and complete contract/abuse-suite verification | #707 | sonnet | apps/gateway, packages/agent, packages/log, plugins/discord | review/tess-m1 | TESS-M1-SEC-001,TESS-M1-SEC-002,TESS-M1-SEC-003,TESS-M1-SEC-004,TESS-M1-SEC-005,TESS-M1-SEC-006,TESS-M1-003,TESS-M1-OBS-001 | 20K | Gate M2 |
+| TESS-M2-001 | not-started | Add Tess roster/profile/service pinned to GPT-5.6 Sol high with fail-fast config and observable effective policy | #708 | codex | packages/mosaic/framework | feat/tess-pi-service | TESS-M1-V | 22K | TESS-PI-001; explicit AC-TESS-03 test |
+| TESS-M2-002 | not-started | Implement durable session identity, inbox/outbox, approval, checkpoint, handoff, compaction and restart recovery | #708 | codex | apps/gateway, packages/agent, packages/db | feat/tess-durable-state | TESS-M2-001 | 38K | TESS-STA-001, TESS-SEC-007..008; recovery TDD |
+| TESS-M2-V | not-started | Clean-host Pi launch plus model/policy status and restart/compaction/duplicate-side-effect verification | #708 | sonnet | apps/gateway/src/__tests__/integration, packages/mosaic/src | review/tess-m2 | TESS-M2-002 | 18K | Gate M3; AC-TESS-03/06 |
+| TESS-M3-001 | not-started | Bind dedicated Tess Discord channel with streaming, threads, attachments, pairing/RBAC and approvals | #709 | codex | plugins/discord, apps/gateway | feat/tess-discord | TESS-M2-V,TESS-M1-SEC-004 | 35K | TESS-DSC-001 |
+| TESS-M3-002 | not-started | Implement `mosaic tess` chat/status/sessions/tree/attach/send/stop/health/recover CLI | #709 | codex | packages/mosaic | feat/tess-cli | TESS-M2-V | 30K | TESS-CLI-001 |
+| TESS-M3-V | not-started | Discord+CLI same-session E2E, denial/approval tests, and operator-flow review | #709 | sonnet | apps/gateway/src/__tests__/integration, plugins/discord, packages/mosaic/src | review/tess-m3 | TESS-M3-001,TESS-M3-002 | 20K | Gate M4 |
+| TESS-M4-001 | not-started | Implement Mos coordination handoff/observe/result contract with authority-boundary tests | #710 | codex | packages/coord, apps/gateway | feat/tess-mos-coordination | TESS-M3-V | 25K | TESS-MOS-001 |
+| TESS-M4-002 | not-started | Implement transitional Hermes runtime/capability adapter | #710 | codex | packages/agent, apps/gateway | feat/tess-hermes-adapter | TESS-M3-V | 40K | TESS-HRM-001; no legacy schema in core contracts |
+| TESS-M4-003 | not-started | Implement memory/retrieval, state/inbox, runtime bootstrap, fleet diagnostics and GitOps plugin foundations | #710 | codex | packages/memory, packages/agent, packages/mosaic | feat/tess-operator-plugins | TESS-M3-V | 40K | TESS-MEM-001, TESS-PLG-001 |
+| TESS-M4-V | not-started | Cross-provider capability, privacy, authority and failure-path qualification | #710 | sonnet | apps/gateway/src/__tests__/integration, packages/agent | review/tess-m4 | TESS-M4-001,TESS-M4-002,TESS-M4-003 | 22K | Gate M5 |
+| TESS-M5-001 | not-started | Implement Matrix/native runtime provider behind common contracts and parity suite | #711 | codex | packages/mosaic, packages/agent | feat/tess-matrix-provider | TESS-M4-V | 30K | TESS-TRN-001 |
+| TESS-M5-002 | not-started | Complete migration inventory, cutover, rollback, retention and deprecation evidence | #711 | sonnet | docs/tess | feat/tess-migration-docs | TESS-M4-V | 18K | TESS-MIG-001 |
+| TESS-M5-003 | not-started | Complete OpenAPI, user/admin/developer/plugin/operations docs and checklist | #711 | codex | docs | feat/tess-docs | TESS-M5-001,TESS-M5-002 | 22K | Documentation hard gate |
+| TESS-M5-V | not-started | Full baseline, contract, integration, Discord/CLI E2E, security review, recovery drill and rollback qualification | #711 | sonnet | apps/gateway, packages/agent, plugins/discord, packages/mosaic | review/tess-final | TESS-M5-003 | 35K | Maps AC-TESS-01..11 to evidence |

docs/tess/THREAT-MODEL.md

@@ -0,0 +1,46 @@
+# Tess Threat Model
+
+## Assets and Trust Boundaries
+
+Assets: operator identity, tenant/project data, agent sessions, fleet control, approvals, credentials, memories, tool outputs, audit evidence, and provider transports.
+
+Trust boundaries: Discord→plugin, CLI→gateway, plugin→gateway service identity, gateway→Pi/provider, Tess→Mos/fleet, Tess→Hermes, MCP→gateway, persistence, and tmux/Matrix transports.
+
+## Threat Matrix
+
+| ID    | Severity | Threat                                                                   | Required control                                                                                                                     | Required verification                                                     |
+| ----- | -------- | ------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------- |
+| TM-01 | critical | Client invokes admin/system command without role                         | Server-side scope/role enforcement in executor; durable approval for privileged/destructive commands                                 | Authenticated non-admin and forged-scope tests deny and audit             |
+| TM-02 | critical | Cross-user/tenant list, attach, send, or terminate by guessed session ID | Owner/tenant binding on every session operation; admin override is explicit and audited                                              | Cross-tenant matrix for REST, WS, CLI, Discord and provider methods       |
+| TM-03 | high     | MCP caller supplies another `userId`                                     | Remove actor IDs from schemas; derive actor/tenant from authenticated context; per-tool scopes                                       | Forged actor/tool calls deny; no victim data returned                     |
+| TM-04 | high     | Discord ingress impersonates user/channel or bypasses gateway auth       | Service-to-service identity, guild/channel/user allowlists, signed/correlated envelope, replay protection                            | Invalid service identity, unlisted IDs, replayed message IDs all deny     |
+| TM-05 | high     | Secrets/PII leak in chat, auth links, tool args, logs, memory, or DB     | Redact before persistence/egress; DM/out-of-band auth flow; short-lived hashed token state; output classification                    | Seeded secret/PII canary absent from durable stores/logs/public channel   |
+| TM-06 | high     | Prompt/tool injection escalates from content to privileged action        | Treat messages/files/tool output as untrusted data; structured proposals only; allowlisted tools; approval binds exact action digest | Injection corpus cannot invoke unapproved tools or alter authority        |
+| TM-07 | high     | Approval forged, replayed, or applied to modified action                 | One-time approval with actor, tenant, action digest, expiry, correlation and consumption record                                      | Forged/replayed/expired/mutated approvals deny and audit                  |
+| TM-08 | medium   | Restart causes message loss or duplicate side effects                    | Durable inbox/outbox/checkpoint; idempotency keys; transactional state transitions; bounded replay                                   | Kill/restart at each state transition; exactly-once effect or safe dedupe |
+| TM-09 | medium   | Session GC/retention crosses tenant/session scope                        | Session/user-scoped GC or separately authorized global retention job                                                                 | GC one session; unrelated logs/memory remain unchanged                    |
+| TM-10 | high     | tmux/Matrix transport target or identity spoofing                        | Exact target/socket binding, peer identity verification, Matrix whoami, authenticated transport metadata                             | Wrong socket/peer/room/identity refuses delivery/attach                   |
+| TM-11 | medium   | Hermes adapter exposes unsupported or broader legacy powers              | Capability negotiation, default deny, normalized scopes, adapter sandbox/timeouts                                                    | Unsupported and over-scoped operations fail closed                        |
+| TM-12 | medium   | Tess competes with Mos or bypasses orchestration gates                   | Authority policy and correlated Mos handoff; no Tess worker-claim capability by default                                              | Coding/decomposition intent produces handoff, not direct claim            |
+
+## Security Invariants
+
+1. Authentication is not authorization; every command/tool/provider operation is authorized server-side.
+2. Actor, tenant, roles, and channel bindings come only from authenticated gateway context.
+3. No client-provided session ID grants ownership or attachment.
+4. No privileged action executes without a matching, unexpired, one-time approval when policy requires it.
+5. Redaction occurs before persistence and before channel egress.
+6. Every externally caused operation is replay-safe and correlated.
+7. Provider capability absence is a denial, not an invitation to shell around it.
+
+## Existing Findings That Block Tess
+
+- Command executor lacks server-side enforcement for declared scopes.
+- Session list/reuse/destroy surfaces are not owner-filtered consistently.
+- MCP schemas accept caller-supplied user identity.
+- Discord plugin lacks a complete authenticated service ingress and user/channel allowlists.
+- Chat/tool persistence lacks mandatory redaction.
+- Sessions/pending Discord output are in-memory and not restart-safe.
+- Session GC currently performs globally scoped promotion.
+
+These are tracked as M1 security prerequisites and must pass independent security review before Tess ingress is enabled.

docs/tess/VERIFICATION-MATRIX.md

@@ -0,0 +1,30 @@
+# Tess Verification Matrix
+
+| Acceptance criterion | Requirements                             | Planned evidence                                                                                                                | Gate             |
+| -------------------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ---------------- |
+| AC-TESS-01           | TESS-PI-001, TESS-DSC-001, TESS-CLI-001  | Discord/CLI same-session integration and streaming E2E                                                                          | M3-V             |
+| AC-TESS-02           | TESS-ARP-001, TESS-CLI-001, TESS-FLT-001 | CLI contract tests for status/sessions/tree/attach/send/stop, typed denial/error snapshots                                      | M3-V             |
+| AC-TESS-03           | TESS-PI-001, TESS-OBS-001                | Clean service launch; status asserts GPT-5.6 Sol, high reasoning and effective tool policy with secret canaries absent          | M2-V, M3-V       |
+| AC-TESS-04           | TESS-MOS-001, TESS-FLT-001               | Authority E2E: coding request creates Mos handoff; safe status runs in Tess; no competing worker claim                          | M4-V             |
+| AC-TESS-05           | TESS-HRM-001                             | Hermes capability contract suite: sessions/stream/send/tree plus Kanban/skills/memory/tools/cron supported-or-denied matrix     | M4-V             |
+| AC-TESS-06           | TESS-STA-001, TESS-SEC-008               | Kill/restart/compaction fault injection across inbox/outbox/checkpoint transitions; duplicate side-effect detector              | M2-V, M5-V       |
+| AC-TESS-07           | TESS-SEC-001..009                        | Threat-model abuse suite: authz, tenant isolation, forged identity/approval, injection, redaction, transport identity, GC scope | M1-V, M3-V, M5-V |
+| AC-TESS-08           | TESS-TRN-001                             | Common provider contract suite against tmux/fleet and Matrix/native; identity and replay tests                                  | M5-V             |
+| AC-TESS-09           | all                                      | `pnpm typecheck`, lint, format, unit/integration/contract/E2E; independent code and security reviews; CI URLs                   | Every milestone  |
+| AC-TESS-10           | TESS-MIG-001                             | Completed capability inventory with native/adapted/deferred/rejected state, owner, cutover/rollback evidence                    | M5-V             |
+| AC-TESS-11           | TESS-PLG-001, TESS-OBS-001               | OpenAPI and user/admin/developer/plugin/ops docs, sitemap links, documentation checklist                                        | M5-V             |
+
+## Security Abuse Suite Minimum
+
+- Role/scope matrix for every command and provider capability.
+- Cross-tenant and cross-user session ID matrix across REST, WS, Discord, CLI, MCP, and providers.
+- Discord service identity, guild/channel/user allowlist, replay, attachment, and mention/DM policy cases.
+- Prompt/tool injection corpus and structured-proposal enforcement.
+- Approval action-digest mutation, replay, expiry, tenant, and actor mismatch cases.
+- Secret/PII canaries through message, attachment, tool args/output, logs, memory, audit, and error paths.
+- Restart fault injection before/after enqueue, provider send, side effect, response persistence, and acknowledgement.
+- Wrong tmux socket/target and Matrix identity/room/replay cases.
+
+## Evidence Rules
+
+Evidence must include command/test name, terminal result, CI run URL, PR/merge reference, environment, and artifact/log location. A worker self-report is not evidence until independently verified.

eslint.config.mjs

@@ -30,6 +30,7 @@ export default tseslint.config(
             'apps/gateway/vitest.config.ts',
             'packages/db/vitest.config.ts',
             'packages/storage/vitest.config.ts',
+            'packages/mosaic/vitest.config.ts',
             'packages/mosaic/__tests__/*.ts',
             'tools/federation-harness/*.ts',
           ],

guides/BOOTSTRAP.md

@@ -15,6 +15,22 @@ This guide covers how to bootstrap a project so AI agents (Claude, Codex, etc.)
 7. Branching/merging is consistent: `branch -> main` via PR with squash-only merges
 8. Steered-autonomy execution is enabled so agents can run end-to-end with escalation-only human intervention
 
+## Agent Host Prerequisites
+
+Agent hosts must provide the Python runtime shape that runtime agents and
+Mosaic automation assume is present.
+
+For Debian/Ubuntu hosts:
+
+```bash
+sudo apt-get update
+# #561: bare python invocations from agents must resolve.
+sudo apt-get install -y python3 python-is-python3
+```
+
+For non-Debian hosts, install the equivalent Python 3 runtime and ensure
+`/usr/bin/python` resolves to `python3` (for example, via a managed symlink).
+
 ## Quick Start
 
 ```bash

packages/db/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mosaicstack/db",
-  "version": "0.0.3",
+  "version": "0.0.4",
   "repository": {
     "type": "git",
     "url": "https://git.mosaicstack.dev/mosaicstack/stack.git",

packages/mosaic/framework/guides/BOOTSTRAP.md

@@ -15,6 +15,22 @@ This guide covers how to bootstrap a project so AI agents (Claude, Codex, etc.)
 7. Branching/merging is consistent: `branch -> main` via PR with squash-only merges
 8. Steered-autonomy execution is enabled so agents can run end-to-end with escalation-only human intervention
 
+## Agent Host Prerequisites
+
+Agent hosts must provide the Python runtime shape that runtime agents and
+Mosaic automation assume is present.
+
+For Debian/Ubuntu hosts:
+
+```bash
+sudo apt-get update
+# #561: bare python invocations from agents must resolve.
+sudo apt-get install -y python3 python-is-python3
+```
+
+For non-Debian hosts, install the equivalent Python 3 runtime and ensure
+`/usr/bin/python` resolves to `python3` (for example, via a managed symlink).
+
 ## Quick Start
 
 ```bash

packages/mosaic/framework/tools/_lib/credentials.sh

@@ -15,13 +15,23 @@
 #
 # After loading, service-specific env vars are exported.
 # Run `load_credentials --help` for details.
+#
+# Resolution order (first match wins):
+#   1. $MOSAIC_CREDENTIALS_FILE (explicit override — never second-guessed)
+#   2. $HOME/.config/mosaic/credentials.json
+#   3. /etc/mosaic/credentials.json (host-level fallback)
+# The /etc fallback exists for HOME-redirected profile environments, where
+# $HOME points at a per-profile directory that has no credentials file.
+# Operators symlink /etc/mosaic/credentials.json to the host's canonical
+# file once, instead of exporting MOSAIC_CREDENTIALS_FILE per invocation.
 
 if [[ -z "${MOSAIC_CREDENTIALS_FILE:-}" ]]; then
-  for _cand in "$HOME/.config/mosaic/credentials.json"; do
+  for _cand in "$HOME/.config/mosaic/credentials.json" "/etc/mosaic/credentials.json"; do
     if [[ -f "$_cand" ]]; then MOSAIC_CREDENTIALS_FILE="$_cand"; break; fi
   done
   : "${MOSAIC_CREDENTIALS_FILE:=$HOME/.config/mosaic/credentials.json}"
 fi
+export MOSAIC_CREDENTIALS_FILE
 
 _mosaic_require_jq() {
   if ! command -v jq &>/dev/null; then

packages/mosaic/framework/tools/fleet/start-agent-session.sh

@@ -114,10 +114,21 @@ MOSAIC_RUNTIME_BIN_PREFIX=$(_build_runtime_bin_prefix)
 # safe single bash token regardless of the name's characters.
 AGENT_NAME_Q=$(printf '%q' "$AGENT_NAME")
 
+# MOSAIC_AGENT_CLASS must ALSO be exported INTO the pane, for the same reason as
+# MOSAIC_AGENT_NAME above: the pane inherits the tmux SERVER environment (not this
+# script's env, and not the systemd unit's EnvironmentFile), so the per-agent class
+# written to agents/<name>.env would otherwise be invisible in-pane. The launcher
+# composes the persona contract from process.env.MOSAIC_AGENT_CLASS at launch
+# (compose-contract -> readPersonaContractBlock); without this export it sees an
+# undefined class and silently injects NO persona contract. %q-quote it so it is a
+# safe single bash token; an empty/unset class %q-quotes to '' and is a harmless
+# no-op downstream (readPersonaContractBlock returns '' for an empty class).
+AGENT_CLASS_Q=$(printf '%q' "${MOSAIC_AGENT_CLASS:-}")
+
 if [ -n "$MOSAIC_RUNTIME_BIN_PREFIX" ]; then
-  PANE_SHELL_SNIPPET="export MOSAIC_AGENT_NAME=${AGENT_NAME_Q}; export PATH=\"${MOSAIC_RUNTIME_BIN_PREFIX}:\${PATH}\"; exec ${MOSAIC_AGENT_COMMAND}"
+  PANE_SHELL_SNIPPET="export MOSAIC_AGENT_NAME=${AGENT_NAME_Q}; export MOSAIC_AGENT_CLASS=${AGENT_CLASS_Q}; export PATH=\"${MOSAIC_RUNTIME_BIN_PREFIX}:\${PATH}\"; exec ${MOSAIC_AGENT_COMMAND}"
 else
-  PANE_SHELL_SNIPPET="export MOSAIC_AGENT_NAME=${AGENT_NAME_Q}; exec ${MOSAIC_AGENT_COMMAND}"
+  PANE_SHELL_SNIPPET="export MOSAIC_AGENT_NAME=${AGENT_NAME_Q}; export MOSAIC_AGENT_CLASS=${AGENT_CLASS_Q}; exec ${MOSAIC_AGENT_COMMAND}"
 fi
 
 mkdir -p "$MOSAIC_AGENT_WORKDIR"

packages/mosaic/framework/tools/fleet/test-start-agent-session.sh

@@ -104,6 +104,7 @@ PATH="$FAKE_BIN:$PATH" \
 MOSAIC_TMUX_SOCKET="$SOCKET3" \
 MOSAIC_AGENT_WORKDIR="$WORKDIR3" \
 MOSAIC_AGENT_RUNTIME="pi" \
+MOSAIC_AGENT_CLASS="code" \
 MOSAIC_RUNTIME_BIN="$FAKE_RUNTIME_BIN" \
 MOSAIC_AGENT_COMMAND="mosaic yolo pi --model openai-codex/gpt-5.5:high" \
 MOSAIC_HEARTBEAT_RUN_DIR="$HB_RUN_DIR3" \
@@ -127,6 +128,18 @@ echo "$all_args" | grep -qF "exec " || fail "pane command does not use exec"
 echo "$all_args" | grep -qF "mosaic yolo pi --model openai-codex/gpt-5.5:high" || \
   fail "pane command does not forward MOSAIC_AGENT_COMMAND with flags intact"
 
+# d) MOSAIC_AGENT_NAME and the per-agent MOSAIC_AGENT_CLASS must BOTH be exported
+#    INTO the pane. The pane inherits the tmux SERVER environment (not this
+#    script's env, nor the systemd unit's EnvironmentFile), so any per-agent var
+#    the launcher needs in-pane must be re-exported in the snippet. CLASS is
+#    load-bearing: the launcher composes the persona contract from
+#    process.env.MOSAIC_AGENT_CLASS, so a missing export silently drops the
+#    persona (regression guard for the A3a pane-propagation gap).
+echo "$all_args" | grep -qF "export MOSAIC_AGENT_NAME=" || \
+  fail "pane command does not export MOSAIC_AGENT_NAME into the pane"
+echo "$all_args" | grep -qF "export MOSAIC_AGENT_CLASS=code" || \
+  fail "pane command does not export MOSAIC_AGENT_CLASS into the pane (persona would silently drop)"
+
 # ── Test 4: when no extra runtime-bin dirs exist, exec still appears ───────────
 TMUX_ARGS_FILE2=$(mktemp)
 FAKE_BIN2=$(mktemp -d)

packages/mosaic/framework/tools/git/detect-platform.sh

@@ -86,7 +86,16 @@ gitea_url_matches_host() {
 
 get_gitea_service_for_host() {
     local host="$1"
-    local cred_file="${MOSAIC_CREDENTIALS_FILE:-$HOME/.config/mosaic/credentials.json}"
+    local cred_file="${MOSAIC_CREDENTIALS_FILE:-}"
+    if [[ -z "$cred_file" ]]; then
+        # Same resolution chain as _lib/credentials.sh: profile HOME, then
+        # host-level /etc only if it exists; neither existing keeps the
+        # $HOME default (matches the lib's final := fallback).
+        cred_file="$HOME/.config/mosaic/credentials.json"
+        if [[ ! -f "$cred_file" && -f /etc/mosaic/credentials.json ]]; then
+            cred_file="/etc/mosaic/credentials.json"
+        fi
+    fi
 
     case "$host" in
         git.mosaicstack.dev)
@@ -231,6 +240,33 @@ get_gitea_login_for_host() {
     return 1
 }
 
+# Validate the current authenticated Gitea user for a resolved Tea login.
+# Tea stores a user name with each login which can become stale after user rename,
+# token rotation, or server migration. Querying /user derives the identity from the
+# active credential instead of trusting that saved name. Callers fall back to the
+# host-scoped API path when this validation fails.
+get_gitea_authenticated_user() {
+    local login_name="$1" response
+
+    command -v tea >/dev/null 2>&1 || return 1
+    response=$(tea api --login "$login_name" /user 2>/dev/null) || return 1
+    TEA_AUTHENTICATED_USER_JSON="$response" python3 - <<'PY'
+import json
+import os
+
+try:
+    user = json.loads(os.environ["TEA_AUTHENTICATED_USER_JSON"])
+except (KeyError, json.JSONDecodeError):
+    raise SystemExit(1)
+
+login = user.get("login") if isinstance(user, dict) else None
+if isinstance(login, str) and login:
+    print(login)
+    raise SystemExit(0)
+raise SystemExit(1)
+PY
+}
+
 get_default_tea_login() {
     local logins_json
 

packages/mosaic/framework/tools/git/issue-assign.sh

@@ -33,7 +33,7 @@ Examples:
     $(basename "$0") -i 42 -l "in-progress" -m "0.2.0"
     $(basename "$0") -i 42 -a @me
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -60,7 +60,7 @@ while [[ $# -gt 0 ]]; do
             shift
             ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2

packages/mosaic/framework/tools/git/issue-create.sh

@@ -12,6 +12,7 @@ TITLE=""
 BODY=""
 LABELS=""
 MILESTONE=""
+INTERACTIVE=false
 
 # get_remote_host and get_gitea_token are provided by detect-platform.sh
 
@@ -66,13 +67,15 @@ Options:
     -b, --body BODY         Issue body/description
     -l, --labels LABELS     Comma-separated labels (e.g., "bug,feature")
     -m, --milestone NAME    Milestone name to assign
+    -i, --interactive       Prompt for missing issue fields
     -h, --help              Show this help message
 
 Examples:
     $(basename "$0") -t "Fix login bug" -l "bug,priority-high"
     $(basename "$0") -t "Add dark mode" -b "Implement theme switching" -m "0.2.0"
+    $(basename "$0") -i
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -94,8 +97,12 @@ while [[ $# -gt 0 ]]; do
             MILESTONE="$2"
             shift 2
             ;;
+        -i|--interactive)
+            INTERACTIVE=true
+            shift
+            ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2
@@ -104,6 +111,13 @@ while [[ $# -gt 0 ]]; do
     esac
 done
 
+if [[ "$INTERACTIVE" == true ]]; then
+    [[ -n "$TITLE" ]] || read -r -p "Issue title: " TITLE
+    [[ -n "$BODY" ]] || read -r -p "Issue body (optional): " BODY || true
+    [[ -n "$LABELS" ]] || read -r -p "Labels, comma-separated (optional): " LABELS || true
+    [[ -n "$MILESTONE" ]] || read -r -p "Milestone (optional): " MILESTONE || true
+fi
+
 if [[ -z "$TITLE" ]]; then
     echo "Error: Title is required (-t)" >&2
     usage
@@ -127,6 +141,11 @@ case "$PLATFORM" in
                 gitea_issue_create_api
                 exit $?
             }
+            if ! get_gitea_authenticated_user "$GITEA_LOGIN_NAME" >/dev/null; then
+                echo "Warning: Tea authenticated-user validation failed (possible stale user/login); trying Gitea API fallback..." >&2
+                gitea_issue_create_api
+                exit $?
+            fi
             REPO_ARGS=(--repo "$REPO_SLUG" --login "$GITEA_LOGIN_NAME")
             CMD=(tea issue create "${REPO_ARGS[@]}" --title "$TITLE")
             [[ -n "$BODY" ]] && CMD+=(--description "$BODY")

packages/mosaic/framework/tools/git/issue-list.sh

@@ -36,7 +36,7 @@ Examples:
     $(basename "$0") -m "0.2.0"               # Issues in milestone 0.2.0
     $(basename "$0") --repo ddk/ai-bma         # List issues from anywhere
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -67,7 +67,7 @@ while [[ $# -gt 0 ]]; do
             shift 2
             ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2

packages/mosaic/framework/tools/git/milestone-create.sh

@@ -37,7 +37,7 @@ Examples:
     $(basename "$0") -t "0.0.1" -d "Pre-MVP Foundation Sprint"
     $(basename "$0") -t "0.1.0" -d "MVP Release" --due "2025-03-01"
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -60,7 +60,7 @@ while [[ $# -gt 0 ]]; do
             shift
             ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2

packages/mosaic/framework/tools/git/pr-create.sh

@@ -86,7 +86,7 @@ Examples:
     $(basename "$0") -i 42 -b "Implements the feature described in #42"
     $(basename "$0") -t "WIP: New feature" --draft
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -125,7 +125,7 @@ while [[ $# -gt 0 ]]; do
             shift
             ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2
@@ -183,6 +183,11 @@ case "$PLATFORM" in
             gitea_pr_create_api
             exit $?
         }
+        if ! get_gitea_authenticated_user "$GITEA_LOGIN_NAME" >/dev/null; then
+            echo "Warning: Tea authenticated-user validation failed (possible stale user/login); trying Gitea API fallback..." >&2
+            gitea_pr_create_api
+            exit $?
+        fi
         REPO_ARGS=(--repo "$REPO_SLUG" --login "$GITEA_LOGIN_NAME")
         CMD=(tea pr create "${REPO_ARGS[@]}" --title "$TITLE")
         [[ -n "$BODY" ]] && CMD+=(--description "$BODY")

packages/mosaic/framework/tools/git/pr-list.sh

@@ -34,7 +34,7 @@ Examples:
     $(basename "$0") -s merged -a username    # Merged PRs by user
     $(basename "$0") --repo ddk/ai-bma         # List PRs from anywhere
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -61,7 +61,7 @@ while [[ $# -gt 0 ]]; do
             shift 2
             ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2

packages/mosaic/framework/tools/git/pr-merge.sh

@@ -35,7 +35,7 @@ Examples:
     $(basename "$0") -n 42 -d                 # Squash merge and delete branch
     $(basename "$0") -n 42 --skip-queue-guard # Skip queue guard wait
 EOF
-    exit 1
+    exit "${1:-1}"
 }
 
 # Parse arguments
@@ -63,7 +63,7 @@ while [[ $# -gt 0 ]]; do
             shift
             ;;
         -h|--help)
-            usage
+            usage 0
             ;;
         *)
             echo "Unknown option: $1" >&2

packages/mosaic/framework/tools/git/test-gitea-login-resolution.sh

@@ -45,6 +45,11 @@ JSON
     exit 0
 fi
 
+if [[ "${1:-}" == "api" ]]; then
+    printf '%s\n' '{"login":"ci-bot"}'
+    exit 0
+fi
+
 printf 'tea %s\n' "$*" >> "$MOSAIC_TEST_LOG"
 if [[ "${MOSAIC_TEA_FAIL_PR_CREATE:-}" == "1" && "$*" == pr\ create* ]]; then
     echo 'GetUserByName: simulated stale login failure' >&2

packages/mosaic/framework/tools/git/test-help-exit-code.sh

@@ -0,0 +1,53 @@
+#!/usr/bin/env bash
+# Regression harness for #701: -h/--help must exit 0, bad args must still exit nonzero.
+#
+# Covers the 7 wrappers whose usage() previously hard-coded `exit 1`, so every
+# --help invocation exited nonzero and logged a phantom isError across fleet lanes.
+# Asserts, per wrapper:
+#   1. `--help` exits 0 and prints usage.
+#   2. `-h` exits 0 and prints usage.
+#   3. A genuine unknown flag still exits nonzero (usage() default path untouched).
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+WRAPPERS=(
+    issue-assign.sh
+    issue-create.sh
+    issue-list.sh
+    milestone-create.sh
+    pr-create.sh
+    pr-list.sh
+    pr-merge.sh
+)
+
+fail=0
+
+for wrapper in "${WRAPPERS[@]}"; do
+    path="$SCRIPT_DIR/$wrapper"
+
+    if ! output=$(bash "$path" --help 2>&1); then
+        echo "FAIL: $wrapper --help exited nonzero" >&2
+        fail=1
+    elif [[ "$output" != Usage:* ]]; then
+        echo "FAIL: $wrapper --help did not print usage" >&2
+        fail=1
+    fi
+
+    if ! bash "$path" -h >/dev/null 2>&1; then
+        echo "FAIL: $wrapper -h exited nonzero" >&2
+        fail=1
+    fi
+
+    if bash "$path" --this-is-not-a-real-flag >/dev/null 2>&1; then
+        echo "FAIL: $wrapper accepted an unknown flag (should have exited nonzero)" >&2
+        fail=1
+    fi
+done
+
+if [[ "$fail" -eq 0 ]]; then
+    echo "help-exit-code regression passed (7/7 wrappers)"
+fi
+
+exit "$fail"

packages/mosaic/framework/tools/git/test-issue-create-body-safety.sh

@@ -55,6 +55,11 @@ JSON
     exit 0
 fi
 
+if [[ "${1:-}" == "api" ]]; then
+    printf '%s\n' '{"login":"ci-bot"}'
+    exit 0
+fi
+
 if [[ "${1:-}" == "issue" && "${2:-}" == "create" ]]; then
     desc=""
     while [[ $# -gt 0 ]]; do

packages/mosaic/framework/tools/git/test-issue-create-interactive-auth.sh

@@ -0,0 +1,88 @@
+#!/usr/bin/env bash
+# Regression harness for #703: interactive issue creation and stale Tea-user fallback.
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+WORK_DIR="${MOSAIC_TEST_WORK_DIR:-$PWD/.mosaic-test-work/issue-create-interactive-auth}"
+REPO_DIR="$WORK_DIR/repo"
+BIN_DIR="$WORK_DIR/bin"
+LOG_FILE="$WORK_DIR/calls.log"
+CREDENTIALS_FILE="$WORK_DIR/credentials.json"
+
+rm -rf "$WORK_DIR"
+mkdir -p "$REPO_DIR" "$BIN_DIR"
+git -C "$REPO_DIR" init -q
+git -C "$REPO_DIR" remote add origin https://git.mosaicstack.dev/mosaicstack/stack.git
+
+cat > "$CREDENTIALS_FILE" <<'JSON'
+{"gitea":{"mosaicstack":{"url":"https://git.mosaicstack.dev","token":"test-token"}}}
+JSON
+
+cat > "$BIN_DIR/tea" <<'SH'
+#!/usr/bin/env bash
+set -euo pipefail
+if [[ "$*" == "login list --output json" ]]; then
+  printf '%s\n' '[{"name":"mosaicstack","url":"https://git.mosaicstack.dev"}]'
+  exit 0
+fi
+if [[ "${1:-}" == "api" ]]; then
+  if [[ "${MOSAIC_TEA_STALE_USER:-0}" == "1" ]]; then
+    echo 'GetUserByName: stale configured user' >&2
+    exit 1
+  fi
+  printf '%s\n' '{"login":"current-user"}'
+  exit 0
+fi
+printf 'tea %s\n' "$*" >> "$MOSAIC_TEST_LOG"
+exit 0
+SH
+
+cat > "$BIN_DIR/curl" <<'SH'
+#!/usr/bin/env bash
+set -euo pipefail
+printf 'curl %s\n' "$*" >> "$MOSAIC_TEST_LOG"
+printf '%s\n' '{"number":703}'
+SH
+chmod +x "$BIN_DIR/tea" "$BIN_DIR/curl"
+
+run_wrapper() {
+  (
+    cd "$REPO_DIR"
+    PATH="$BIN_DIR:$PATH" \
+    MOSAIC_CREDENTIALS_FILE="$CREDENTIALS_FILE" \
+    MOSAIC_TEST_LOG="$LOG_FILE" \
+    "$@"
+  )
+}
+
+: > "$LOG_FILE"
+printf 'Interactive title\nInteractive body\nlabel-a,label-b\nM1\n' | run_wrapper "$SCRIPT_DIR/issue-create.sh" -i >/dev/null
+
+grep -q -- 'tea issue create --repo mosaicstack/stack --login mosaicstack --title Interactive title --description Interactive body --labels label-a,label-b --milestone M1' "$LOG_FILE"
+
+# Explicit values take precedence in interactive mode: no title input is
+# supplied, but the wrapper still creates the issue with the explicit title.
+: > "$LOG_FILE"
+printf '\n\n\n' | run_wrapper "$SCRIPT_DIR/issue-create.sh" -i -t 'Explicit title' >/dev/null
+grep -q -- 'tea issue create --repo mosaicstack/stack --login mosaicstack --title Explicit title' "$LOG_FILE"
+
+: > "$LOG_FILE"
+run_wrapper env MOSAIC_TEA_STALE_USER=1 "$SCRIPT_DIR/issue-create.sh" -t 'Fallback title' -b 'Fallback body' >/dev/null 2>"$WORK_DIR/issue-stderr"
+grep -q -- 'curl .*https://git.mosaicstack.dev/api/v1/repos/mosaicstack/stack/issues' "$LOG_FILE"
+grep -q -- 'Tea authenticated-user validation failed' "$WORK_DIR/issue-stderr"
+if grep -q -- 'tea issue create' "$LOG_FILE"; then
+  echo 'FAIL: issue-create invoked Tea mutation after stale-user validation failed' >&2
+  exit 1
+fi
+
+: > "$LOG_FILE"
+run_wrapper env MOSAIC_TEA_STALE_USER=1 "$SCRIPT_DIR/pr-create.sh" -t 'PR fallback' -H feature/wrapfix >/dev/null 2>"$WORK_DIR/pr-stderr"
+grep -q -- 'curl .*https://git.mosaicstack.dev/api/v1/repos/mosaicstack/stack/pulls' "$LOG_FILE"
+grep -q -- 'Tea authenticated-user validation failed' "$WORK_DIR/pr-stderr"
+if grep -q -- 'tea pr create' "$LOG_FILE"; then
+  echo 'FAIL: pr-create invoked Tea mutation after stale-user validation failed' >&2
+  exit 1
+fi
+
+echo 'issue-create interactive/auth regression harness passed'

packages/mosaic/framework/tools/tmux/README.md

@@ -87,6 +87,10 @@ message crosses the wire as base64 (`-b`) to avoid all shell-quoting hazards.
 
 - `agent-send.sh` — inter-agent wrapper (preamble + local/remote dispatch).
 - `send-message.sh` — low-level reliable single-pane submitter (`-b` base64 input).
+- `auto-submit-drafts.sh` — watchdog that flushes stable unsubmitted prompt
+  drafts on a coordinator pane (default target `mos-claude`); run it as a
+  long-lived process alongside the coordinator session.
+- `agent-send.test.sh` — regression + grammar lock for `agent-send.sh`.
 - `test-send-message-socket.sh` — smoke test for named-socket isolation.
 
 ## Distribution

packages/mosaic/framework/tools/tmux/auto-submit-drafts.sh

@@ -0,0 +1,80 @@
+#!/usr/bin/env bash
+# auto-submit-drafts.sh — watchdog for Claude Code panes that receive channel
+# messages but leave them as unsubmitted prompt drafts. Intended for Mos only.
+set -uo pipefail
+
+TARGET="${1:-mos-claude}"
+INTERVAL="${INTERVAL:-2}"
+STABLE_SECONDS="${STABLE_SECONDS:-4}"
+LOG_PREFIX="[auto-submit-drafts:$TARGET]"
+
+last_prompt=""
+first_seen=0
+
+prompt_text() {
+  tmux capture-pane -t "$TARGET" -p 2>/dev/null | python3 -c '
+import sys, re
+lines = sys.stdin.read().splitlines()
+idx = None
+for i in range(len(lines)-1, -1, -1):
+    if "❯" in lines[i]:
+        idx = i
+        break
+if idx is None:
+    raise SystemExit
+parts = []
+after = lines[idx].split("❯", 1)[1]
+parts.append(after)
+for line in lines[idx+1:]:
+    # Stop at Claude Code separator/border lines.
+    if "─" in line or "╰" in line or "╭" in line:
+        break
+    s = line.replace("\u00a0", " ")
+    s = re.sub(r"[\x00-\x1f\x7f]", "", s).strip()
+    if s:
+        parts.append(s)
+text = " ".join(parts).replace("\u00a0", " ")
+text = re.sub(r"[\x00-\x1f\x7f]", "", text).strip()
+print(text)
+'
+}
+
+while true; do
+  if ! tmux has-session -t "$TARGET" 2>/dev/null; then
+    echo "$LOG_PREFIX target missing; waiting" >&2
+    sleep "$INTERVAL"
+    last_prompt=""
+    first_seen=0
+    continue
+  fi
+
+  current="$(prompt_text || true)"
+  now="$(date +%s)"
+
+  if [[ -z "$current" ]]; then
+    last_prompt=""
+    first_seen=0
+    sleep "$INTERVAL"
+    continue
+  fi
+
+  if [[ "$current" != "$last_prompt" ]]; then
+    last_prompt="$current"
+    first_seen="$now"
+    sleep "$INTERVAL"
+    continue
+  fi
+
+  age=$(( now - first_seen ))
+  if (( age >= STABLE_SECONDS )); then
+    echo "$LOG_PREFIX submitting stable draft after ${age}s: ${current:0:120}" >&2
+    tmux send-keys -t "$TARGET" C-j
+    sleep 0.8
+    tmux send-keys -t "$TARGET" C-m
+    sleep 2
+    last_prompt=""
+    first_seen=0
+  else
+    sleep "$INTERVAL"
+  fi
+done

packages/mosaic/framework/tools/tmux/send-message.sh

@@ -77,10 +77,20 @@ snippet=$(printf '%s' "$MSG" | tr '\n' ' ' | tr -s ' ' | sed 's/[^[:print:]]//g'
 
 # 1) Paste the body as a bracketed paste so multi-line content does not submit
 #    line-by-line. load-buffer/paste-buffer is far safer than `send-keys -l`.
-printf '%s' "$MSG" | "${tmux_cmd[@]}" load-buffer -b __mosaic_send -
+# Buffer name MUST be unique per invocation: concurrent senders on the shared
+# tmux server race a fixed name (load overwrites load, -d deletes underneath),
+# cross-delivering or dropping messages — bit the fleet on the 2026-07-09
+# simultaneous restart (briefs swapped between sessions).
+BUF="__mosaic_send_$$_$(date +%s%N)"
+printf '%s' "$MSG" | "${tmux_cmd[@]}" load-buffer -b "$BUF" -
 # -p = bracketed paste when the client supports it; fall back if not.
-"${tmux_cmd[@]}" paste-buffer -d -p -b __mosaic_send -t "$EFFECTIVE_TARGET" 2>/dev/null \
-  || "${tmux_cmd[@]}" paste-buffer -d -b __mosaic_send -t "$EFFECTIVE_TARGET"
+"${tmux_cmd[@]}" paste-buffer -d -p -b "$BUF" -t "$EFFECTIVE_TARGET" 2>/dev/null \
+  || "${tmux_cmd[@]}" paste-buffer -d -b "$BUF" -t "$EFFECTIVE_TARGET" \
+  || "${tmux_cmd[@]}" delete-buffer -b "$BUF" 2>/dev/null
+# ^ -d deletes the buffer only on a SUCCESSFUL paste; if both attempts fail
+#   (e.g. the target vanished since the liveness check), delete explicitly —
+#   named buffers are exempt from tmux's buffer-limit eviction, so orphans
+#   would otherwise accumulate forever.
 sleep 0.5
 
 # 2) Submit, then verify; flush with another Enter if it is still a draft.

packages/mosaic/framework/tools/tmux/test-send-message-socket.sh

@@ -47,4 +47,32 @@ if capture_default | grep -qF "agent socket hello"; then
   fail "agent-send.sh leaked named-socket message to default tmux server"
 fi
 
+# Concurrency: parallel senders on one server must not cross-deliver or drop.
+# Locks the unique-per-invocation paste buffer (a fixed buffer name raced:
+# load overwrote load, -d deleted underneath — messages swapped between panes).
+CONC_N=5
+for i in $(seq 1 "$CONC_N"); do
+  tmux -L "$SOCKET" new-session -d -s "conc-$i" -c "$TMPDIR" 'bash --noprofile --norc -i'
+done
+pids=()
+for i in $(seq 1 "$CONC_N"); do
+  "$SEND_MESSAGE" -L "$SOCKET" -t "=conc-$i" -m "CONCPAYLOAD-${i}-END" >/dev/null &
+  pids+=($!)
+done
+for pid in "${pids[@]}"; do
+  wait "$pid" || fail "concurrent send-message.sh invocation exited non-zero"
+done
+sleep 0.2
+for i in $(seq 1 "$CONC_N"); do
+  pane=$(tmux -L "$SOCKET" capture-pane -t "=conc-$i:0.0" -p)
+  printf '%s' "$pane" | grep -qF "CONCPAYLOAD-${i}-END" \
+    || fail "concurrent send dropped payload for pane conc-$i"
+  for j in $(seq 1 "$CONC_N"); do
+    [ "$j" = "$i" ] && continue
+    if printf '%s' "$pane" | grep -qF "CONCPAYLOAD-${j}-END"; then
+      fail "concurrent send cross-delivered payload $j to pane conc-$i"
+    fi
+  done
+done
+
 echo "ok - named tmux socket send tools"

packages/mosaic/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mosaicstack/mosaic",
-  "version": "0.0.45",
+  "version": "0.0.48",
   "repository": {
     "type": "git",
     "url": "https://git.mosaicstack.dev/mosaicstack/stack.git",

packages/mosaic/src/commands/fleet-personas.ts

@@ -234,6 +234,21 @@ export async function resolvePersona(
     extractClassesFromDir(rolesDir),
     extractClassesFromDir(overrideDir),
   ]);
+  return resolvePersonaFrom(klass, { rolesDir, overrideDir, base, over });
+}
+
+/**
+ * Resolve a single class against ALREADY-EXTRACTED layer maps. Callers that
+ * resolve many classes against the same two directories (e.g. provisioning a
+ * full roster) should {@link extractClassesFromDir} each dir ONCE and reuse the
+ * result here, rather than paying a full directory re-scan per class. Precedence
+ * is identical to {@link resolvePersona}: override layer wins, then baseline.
+ */
+export async function resolvePersonaFrom(
+  klass: string,
+  layers: { rolesDir: string; overrideDir: string; base: DirClasses; over: DirClasses },
+): Promise<PersonaResolution | null> {
+  const { rolesDir, overrideDir, base, over } = layers;
 
   const fromLayer = async (
     dir: string,

packages/mosaic/src/commands/fleet-provision.spec.ts

@@ -0,0 +1,270 @@
+import { access, mkdir, mkdtemp, rm, writeFile } from 'node:fs/promises';
+import { constants } from 'node:fs';
+import { tmpdir } from 'node:os';
+import { dirname, join, resolve } from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+import { loadFleetRoster } from './fleet.js';
+import { generateRoster, runProvision } from './fleet-provision.js';
+import { loadProfile } from './fleet-profiles.js';
+
+// These are INTEGRATION tests: each exercises real filesystem I/O — scanning the
+// committed framework/fleet persona library, rendering YAML, writing to a temp
+// mosaicHome, and round-tripping through the real roster parser. On a heavily
+// contended CI runner (the whole monorepo's suites run in parallel) that genuine
+// I/O can exceed vitest's 5s default even though it completes in ~400ms locally.
+// Give the legitimately I/O-bound work generous headroom so CI is deterministic.
+vi.setConfig({ testTimeout: 30_000 });
+
+// The real, committed library: packages/mosaic/src/commands -> framework/fleet.
+const frameworkFleet = resolve(
+  dirname(fileURLToPath(import.meta.url)),
+  '..',
+  '..',
+  'framework',
+  'fleet',
+);
+const rolesDir = join(frameworkFleet, 'roles');
+const profilesDir = join(frameworkFleet, 'profiles');
+
+/** A fresh temp mosaicHome whose fleet/ dir is empty (for write-path tests). */
+async function freshMosaicHome(): Promise<string> {
+  const home = await mkdtemp(join(tmpdir(), 'mosaic-provision-'));
+  await mkdir(join(home, 'fleet'), { recursive: true });
+  return home;
+}
+
+async function fileExists(path: string): Promise<boolean> {
+  try {
+    await access(path, constants.F_OK);
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+describe('provision software-delivery (floor, default)', () => {
+  it('materializes only the floor seats with correct flags + valid scaffold', async () => {
+    const profile = await loadProfile('software-delivery', { profilesDir, rolesDir });
+    const { seats, yaml } = await generateRoster(profile, { profilesDir, rolesDir });
+
+    // Floor is orchestrator + enhancer.
+    expect(seats.map((s) => s.name).sort()).toEqual(['enhancer', 'orchestrator']);
+    const orch = seats.find((s) => s.name === 'orchestrator');
+    const enh = seats.find((s) => s.name === 'enhancer');
+    // RULE 2: floor + lead get persistent_persona.
+    expect(orch?.persistentPersona).toBe(true);
+    expect(enh?.persistentPersona).toBe(true);
+    // RULE 3: floor/lead do NOT reset between tasks.
+    expect(orch?.resetBetweenTasks).toBeUndefined();
+    expect(enh?.resetBetweenTasks).toBeUndefined();
+    // RULE 4: default runtime claude.
+    expect(orch?.runtime).toBe('claude');
+
+    // Scaffold round-trips through the real parser.
+    expect(yaml).toContain('version: 1');
+    expect(yaml).toContain('transport: tmux');
+    expect(yaml).toContain('socket_name: mosaic-fleet');
+  });
+});
+
+describe('provision --full', () => {
+  it('expands the entire roster, including multiplicity, deterministically', async () => {
+    const profile = await loadProfile('software-delivery', { profilesDir, rolesDir });
+    const { seats } = await generateRoster(profile, { full: true, profilesDir, rolesDir });
+
+    const names = seats.map((s) => s.name);
+    // code multiplicity 2 -> code0/code1 (RULE 1).
+    expect(names).toContain('code0');
+    expect(names).toContain('code1');
+    expect(names).not.toContain('code');
+    // Singleton seats keep the bare class name.
+    expect(names).toContain('planner');
+    expect(names).toContain('merge-gate');
+
+    // Deterministic ordering: profile roster order, multiplicity expanded inline.
+    const codeIdx0 = names.indexOf('code0');
+    expect(names[codeIdx0 + 1]).toBe('code1');
+
+    // RULE 3: a non-floor, non-lead execution seat resets between tasks.
+    const code0 = seats.find((s) => s.name === 'code0');
+    expect(code0?.resetBetweenTasks).toBe(true);
+    expect(code0?.persistentPersona).toBeUndefined();
+
+    // Seat count == sum of multiplicities.
+    const expected = profile.roster.reduce((n, e) => n + e.multiplicity, 0);
+    expect(seats.length).toBe(expected);
+  });
+});
+
+describe('generated roster round-trips through the real parser', () => {
+  it('feeds generated YAML back through loadFleetRoster (key correctness proof)', async () => {
+    const home = await freshMosaicHome();
+    try {
+      const result = await runProvision('software-delivery', {
+        mosaicHome: home,
+        profilesDir,
+        rolesDir,
+        full: true,
+        write: true,
+      });
+      expect(result.wrote).toBe(join(home, 'fleet', 'roster.yaml'));
+
+      const parsed = await loadFleetRoster(result.wrote!);
+      // It parses with no error and carries every seat.
+      const profile = await loadProfile('software-delivery', { profilesDir, rolesDir });
+      const expected = profile.roster.reduce((n, e) => n + e.multiplicity, 0);
+      expect(parsed.agents.length).toBe(expected);
+      // Classes survive the round-trip.
+      expect(parsed.agents.some((a) => a.className === 'orchestrator')).toBe(true);
+      expect(parsed.agents.filter((a) => a.className === 'code').length).toBe(2);
+      // reports_to must NOT have been emitted (parser rejects unknown keys).
+      expect(result.yaml).not.toContain('reports_to');
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+});
+
+describe('override-aware persona validation', () => {
+  let dir: string;
+  beforeEach(async () => {
+    dir = await mkdtemp(join(tmpdir(), 'mosaic-provision-ov-'));
+  });
+  afterEach(async () => {
+    await rm(dir, { recursive: true, force: true });
+  });
+
+  it('resolves a user-added (roles.local-only) persona without a false unresolved error', async () => {
+    const overrideDir = join(dir, 'roles.local');
+    const customProfilesDir = join(dir, 'profiles');
+    await mkdir(overrideDir, { recursive: true });
+    await mkdir(customProfilesDir, { recursive: true });
+    // A brand-new class that exists ONLY in roles.local.
+    await writeFile(
+      join(overrideDir, 'widget-maker.md'),
+      '# widget-maker\n\nThe widget-maker persona (`class: widget-maker`).\n',
+    );
+    await writeFile(
+      join(customProfilesDir, 'custom.yaml'),
+      [
+        'id: custom',
+        'title: Custom',
+        'description: a custom system',
+        'lead: widget-maker',
+        'floor: [widget-maker]',
+        'roster:',
+        '  - class: widget-maker',
+        '',
+      ].join('\n'),
+    );
+
+    const result = await runProvision('custom', {
+      mosaicHome: dir,
+      profilesDir: customProfilesDir,
+      // Point baseline rolesDir at a missing dir so resolution depends on override.
+      rolesDir: join(dir, 'no-baseline'),
+      overrideDir,
+    });
+    expect(result.yaml).toContain('class: widget-maker');
+    // It resolved from the override layer.
+    // (generateRoster records personaLayer; the seat is present.)
+    expect(result.summary).toContain('persona=override');
+  });
+
+  it('FAILS with a clear message when a profile references a bogus class', async () => {
+    const customProfilesDir = join(dir, 'profiles');
+    await mkdir(customProfilesDir, { recursive: true });
+    await writeFile(
+      join(customProfilesDir, 'bogus.yaml'),
+      [
+        'id: bogus',
+        'title: Bogus',
+        'description: bad system',
+        'lead: orchestrator',
+        'floor: [orchestrator]',
+        'roster:',
+        '  - class: orchestrator',
+        '  - class: not-a-real-persona-xyz',
+        '    reports_to: orchestrator',
+        '',
+      ].join('\n'),
+    );
+    await expect(
+      runProvision('bogus', {
+        mosaicHome: dir,
+        profilesDir: customProfilesDir,
+        rolesDir,
+        overrideDir: join(dir, 'roles.local'),
+      }),
+    ).rejects.toThrow(/not-a-real-persona-xyz|not a known persona class/);
+  });
+});
+
+describe('--write protection', () => {
+  it('refuses to clobber an existing roster.yaml without --force', async () => {
+    const home = await freshMosaicHome();
+    try {
+      const rosterPath = join(home, 'fleet', 'roster.yaml');
+      await writeFile(rosterPath, 'version: 1\n# operator customizations\n');
+      await expect(
+        runProvision('software-delivery', { mosaicHome: home, profilesDir, rolesDir, write: true }),
+      ).rejects.toThrow(/Refusing to overwrite/);
+      // The original file is untouched.
+      const { readFile } = await import('node:fs/promises');
+      expect(await readFile(rosterPath, 'utf8')).toContain('operator customizations');
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('--write --force overwrites an existing roster', async () => {
+    const home = await freshMosaicHome();
+    try {
+      const rosterPath = join(home, 'fleet', 'roster.yaml');
+      await writeFile(rosterPath, 'version: 1\n# old\n');
+      const result = await runProvision('software-delivery', {
+        mosaicHome: home,
+        profilesDir,
+        rolesDir,
+        write: true,
+        force: true,
+      });
+      expect(result.wrote).toBe(rosterPath);
+      const parsed = await loadFleetRoster(rosterPath);
+      expect(parsed.agents.length).toBeGreaterThan(0);
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('--write to a fresh mosaicHome creates the roster file', async () => {
+    const home = await freshMosaicHome();
+    try {
+      const result = await runProvision('software-delivery', {
+        mosaicHome: home,
+        profilesDir,
+        rolesDir,
+        write: true,
+      });
+      expect(await fileExists(result.wrote!)).toBe(true);
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('dry run (no --write) writes nothing', async () => {
+    const home = await freshMosaicHome();
+    try {
+      const result = await runProvision('software-delivery', {
+        mosaicHome: home,
+        profilesDir,
+        rolesDir,
+      });
+      expect(result.wrote).toBeUndefined();
+      expect(await fileExists(join(home, 'fleet', 'roster.yaml'))).toBe(false);
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+});

packages/mosaic/src/commands/fleet-provision.ts

@@ -0,0 +1,406 @@
+/**
+ * `mosaic fleet provision --profile <id>` — turn a declared SYSTEM TYPE (a
+ * profile) into a concrete fleet roster (North Star H3).
+ *
+ * A profile (fleet/profiles/<id>.yaml) is a DECLARATIVE mapping from a system
+ * type to a persona roster + org topology (H2). This command MATERIALIZES that
+ * declaration into the concrete `roster.yaml` shape the live fleet consumes — the
+ * same shape `fleet.ts` parses (version/transport/tmux/defaults/runtimes/agents).
+ *
+ * DRY-RUN-FIRST + REVIEWABLE: with no --write it prints the roster it WOULD
+ * generate plus a topology summary and writes nothing. --write persists it, and
+ * REFUSES to clobber an existing roster.yaml without --force (protects operator
+ * customizations).
+ *
+ * DRY: profile parsing/validation is reused wholesale from fleet-profiles.ts
+ * (loadProfile/validateProfile) and persona resolution from fleet-personas.ts
+ * (resolvePersona, override-aware). This module owns ONLY the profile→roster
+ * generation policy, documented inline below so each default is reviewable.
+ */
+
+import { access, mkdir, writeFile } from 'node:fs/promises';
+import { constants } from 'node:fs';
+import { homedir } from 'node:os';
+import { dirname, join } from 'node:path';
+import type { Command } from 'commander';
+import YAML from 'yaml';
+import {
+  loadProfile,
+  validateProfile,
+  type FleetProfile,
+  type ProfileRosterEntry,
+  defaultProfilesDir,
+  defaultRolesDir,
+  listPersonaClassesWithOverrides,
+} from './fleet-profiles.js';
+import {
+  defaultOverrideDir,
+  extractClassesFromDir,
+  resolvePersonaFrom,
+  type PersonaLayer,
+} from './fleet-personas.js';
+
+function defaultMosaicHome(): string {
+  return process.env['MOSAIC_HOME'] ?? join(homedir(), '.config', 'mosaic');
+}
+
+// ---------------------------------------------------------------------------
+// GENERATION RULES — each default below is intentionally simple and documented
+// so a reviewer can ratify or override the policy. See the PR body for the open
+// runtime-per-class question (RULE 4).
+// ---------------------------------------------------------------------------
+
+/**
+ * RULE 4 — Runtime assignment policy (THE one open design question).
+ *
+ * Default: EVERY seat → `runtime: claude`. Claude runs every persona, so it is
+ * the safe universal default and guarantees a structurally-valid, launchable
+ * roster regardless of how the policy ultimately lands. We deliberately do NOT
+ * hardcode pi / gpt-5.5 per class here. The live roster today runs coders on
+ * pi + openai-codex/gpt-5.5:high — whether provisioning should encode a
+ * class→runtime/model map (and WHERE: the profile schema vs a separate
+ * runtime-policy file) is flagged in the PR body for ratification.
+ *
+ * Centralized so changing the policy is a ONE-edit change. If a future profile
+ * entry (or persona) declares a runtime/model preference, honor it here; until
+ * then everything defaults to claude.
+ */
+export const DEFAULT_RUNTIME = 'claude';
+
+/** Result of applying the runtime policy to one seat. */
+interface RuntimeChoice {
+  runtime: string;
+  modelHint?: string;
+}
+
+/**
+ * The single centralized runtime-policy function. Today it returns the universal
+ * `claude` default for every seat. To encode a class→runtime/model map later,
+ * edit ONLY this function (and/or extend the profile schema and read it here).
+ */
+export function resolveSeatRuntime(
+  _klass: string,
+  _isFloor: boolean,
+  _isLead: boolean,
+): RuntimeChoice {
+  return { runtime: DEFAULT_RUNTIME };
+}
+
+/** One generated seat, fully resolved for emission + topology display. */
+export interface GeneratedSeat {
+  name: string;
+  className: string;
+  runtime: string;
+  modelHint?: string;
+  persistentPersona?: boolean;
+  resetBetweenTasks?: boolean;
+  /** Topology edge from the profile (NOT emitted to roster.yaml — see RULE 6). */
+  reportsTo?: string;
+  /** Which persona layer the class resolved from (baseline/override). */
+  personaLayer: PersonaLayer;
+}
+
+export interface GenerateRosterResult {
+  seats: GeneratedSeat[];
+  /** The roster.yaml text (parser-valid, drop-in). */
+  yaml: string;
+}
+
+export interface ProvisionOptions {
+  /** Materialize the entire profile roster (multiplicity expanded). */
+  full?: boolean;
+  mosaicHome?: string;
+  /** Test overrides — mirror fleet-profiles LoadProfilesOptions. */
+  profilesDir?: string;
+  rolesDir?: string;
+  overrideDir?: string;
+}
+
+function resolveDirs(opts: ProvisionOptions): {
+  mosaicHome: string;
+  profilesDir: string;
+  rolesDir: string;
+  overrideDir: string;
+} {
+  const mosaicHome = opts.mosaicHome ?? defaultMosaicHome();
+  return {
+    mosaicHome,
+    profilesDir: opts.profilesDir ?? defaultProfilesDir(mosaicHome),
+    rolesDir: opts.rolesDir ?? defaultRolesDir(mosaicHome),
+    overrideDir: opts.overrideDir ?? defaultOverrideDir(mosaicHome),
+  };
+}
+
+/**
+ * RULE 1 — Seat naming. multiplicity 1 → name = class (e.g. `planner`).
+ * multiplicity N>1 → `<class>0`,`<class>1`,… (e.g. `code` ×2 → code0/code1).
+ * Names are deterministic, following profile roster order.
+ */
+function seatNames(entry: ProfileRosterEntry): string[] {
+  if (entry.multiplicity <= 1) return [entry.class];
+  return Array.from({ length: entry.multiplicity }, (_, i) => `${entry.class}${i}`);
+}
+
+/**
+ * Generate the concrete seats + roster.yaml for a validated profile.
+ *
+ * Seat selection:
+ *   --full  → the ENTIRE profile roster, multiplicity expanded.
+ *   default → ONLY the `floor` classes (the always-staffed minimum), matching
+ *             the profile note "two-agent floor always staffed; every other seat
+ *             added on demand."
+ *
+ * Per-seat flags:
+ *   RULE 2 persistent_persona: true for floor classes AND the lead; else omitted.
+ *   RULE 3 reset_between_tasks: true for non-floor, non-lead execution seats;
+ *           floor/lead omit it (mirrors today's coders resetting while the
+ *           orchestrator/enhancer do not).
+ *   RULE 4 runtime: via resolveSeatRuntime (defaults claude).
+ *   RULE 6 reports_to: tracked on the seat for the topology summary but NOT
+ *           emitted to roster.yaml — the fleet.ts parser rejects unknown agent
+ *           keys, so writing reports_to would break round-trip. Confirmed against
+ *           normalizeAgent's allow-list in fleet.ts.
+ *
+ * Persona resolution: every emitted class is resolved override-aware via
+ * resolvePersona so we can (a) record the layer for the summary and (b) refuse
+ * to emit a roster that references a nonexistent persona.
+ */
+export async function generateRoster(
+  profile: FleetProfile,
+  opts: ProvisionOptions,
+): Promise<GenerateRosterResult> {
+  const { rolesDir, overrideDir } = resolveDirs(opts);
+  const floor = new Set(profile.floor);
+  const lead = profile.lead;
+
+  const selected: ProfileRosterEntry[] = opts.full
+    ? profile.roster
+    : profile.roster.filter((e) => floor.has(e.class));
+
+  // Scan the persona directories ONCE, then resolve every roster entry against
+  // the in-memory maps. resolvePersona() would otherwise re-scan both dirs per
+  // entry — O(entries × files) redundant reads that push --full provisioning
+  // past the test timeout on slow/contended filesystems.
+  const [base, over] = await Promise.all([
+    extractClassesFromDir(rolesDir),
+    extractClassesFromDir(overrideDir),
+  ]);
+
+  const seats: GeneratedSeat[] = [];
+  for (const entry of selected) {
+    const isFloor = floor.has(entry.class);
+    const isLead = entry.class === lead;
+
+    const resolved = await resolvePersonaFrom(entry.class, { rolesDir, overrideDir, base, over });
+    if (!resolved) {
+      // Defensive: validateProfile already guards this, but a class can resolve
+      // for membership yet have no readable file. Fail loudly rather than emit a
+      // roster pointing at a persona we cannot load.
+      throw new Error(
+        `Cannot provision: roster class "${entry.class}" does not resolve to a readable persona.`,
+      );
+    }
+
+    const runtimeChoice = resolveSeatRuntime(entry.class, isFloor, isLead);
+    for (const name of seatNames(entry)) {
+      const seat: GeneratedSeat = {
+        name,
+        className: entry.class,
+        runtime: runtimeChoice.runtime,
+        personaLayer: resolved.layer,
+      };
+      if (runtimeChoice.modelHint) seat.modelHint = runtimeChoice.modelHint;
+      if (isFloor || isLead) seat.persistentPersona = true;
+      if (!isFloor && !isLead) seat.resetBetweenTasks = true;
+      if (entry.reportsTo) seat.reportsTo = entry.reportsTo;
+      seats.push(seat);
+    }
+  }
+
+  if (seats.length === 0) {
+    throw new Error(
+      `Profile "${profile.id}" produced no seats. ` +
+        (opts.full ? 'Its roster is empty.' : 'No floor seats are defined — try --full.'),
+    );
+  }
+
+  return { seats, yaml: renderRosterYaml(seats) };
+}
+
+/**
+ * RULE 5 — Standard roster scaffolding. We emit the same generic, non-personal
+ * scaffold the committed example presets use (socket_name: mosaic-fleet,
+ * holder_session: _holder, working_directory: ~, claude + pi runtimes) so the
+ * output is a drop-in valid roster. No operator-personal data is copied.
+ *
+ * Built via the `yaml` lib (same serializer the parser uses) so the result
+ * round-trips. reports_to is intentionally NOT included on agents (RULE 6).
+ */
+function renderRosterYaml(seats: GeneratedSeat[]): string {
+  const agents = seats.map((s) => {
+    const a: Record<string, unknown> = {
+      name: s.name,
+      runtime: s.runtime,
+      class: s.className,
+    };
+    if (s.persistentPersona) a['persistent_persona'] = true;
+    if (s.modelHint) a['model_hint'] = s.modelHint;
+    if (s.resetBetweenTasks) a['reset_between_tasks'] = true;
+    return a;
+  });
+
+  const doc = {
+    version: 1,
+    transport: 'tmux',
+    tmux: { socket_name: 'mosaic-fleet', holder_session: '_holder' },
+    defaults: { working_directory: '~' },
+    runtimes: {
+      claude: { reset_command: '/clear' },
+      pi: { reset_command: '/new' },
+    },
+    agents,
+  };
+
+  return YAML.stringify(doc);
+}
+
+// ---------------------------------------------------------------------------
+// Validation — reuse fleet-profiles.validateProfile (override-aware classes) and
+// name any unresolved class clearly. Never generate a roster referencing a
+// nonexistent persona.
+// ---------------------------------------------------------------------------
+
+/**
+ * Validate the profile against the override-aware persona library. Throws with a
+ * clear, class-naming message if any referenced class is unresolved.
+ */
+export async function validateProfileForProvision(
+  profile: FleetProfile,
+  opts: ProvisionOptions,
+): Promise<void> {
+  const { rolesDir, overrideDir } = resolveDirs(opts);
+  const validClasses = await listPersonaClassesWithOverrides(rolesDir, overrideDir);
+  const problems = validateProfile(profile, validClasses);
+  if (problems.length > 0) {
+    throw new Error(
+      `Profile "${profile.id}" is invalid; cannot provision:\n  - ${problems.join('\n  - ')}`,
+    );
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Topology summary (printed in dry-run and after write).
+// ---------------------------------------------------------------------------
+
+function formatTopologySummary(seats: GeneratedSeat[]): string {
+  const lines: string[] = [];
+  lines.push(`Topology (${seats.length} seat(s)):`);
+  for (const s of seats) {
+    const reports = s.reportsTo ? `reports_to=${s.reportsTo}` : 'reports_to=- (lead)';
+    lines.push(
+      `  - ${s.name}\tclass=${s.className}\truntime=${s.runtime}\t${reports}\tpersona=${s.personaLayer}`,
+    );
+  }
+  return lines.join('\n');
+}
+
+// ---------------------------------------------------------------------------
+// CLI wiring — mirror registerFleetProfileCommand / registerFleetPersonaCommand.
+// ---------------------------------------------------------------------------
+
+export interface ProvisionRunResult {
+  yaml: string;
+  summary: string;
+  wrote?: string;
+}
+
+/**
+ * Core provision flow shared by the CLI: load + validate the profile, generate
+ * the roster, optionally write it. Returns the artifacts for printing/testing.
+ */
+export async function runProvision(
+  profileId: string,
+  opts: ProvisionOptions & { write?: boolean; force?: boolean },
+): Promise<ProvisionRunResult> {
+  const dirs = resolveDirs(opts);
+  const profile = await loadProfile(profileId, {
+    mosaicHome: dirs.mosaicHome,
+    profilesDir: dirs.profilesDir,
+    rolesDir: dirs.rolesDir,
+    overrideDir: dirs.overrideDir,
+  });
+  // loadProfile already validates, but re-run with our explicit error wording so
+  // an unresolved class is named clearly even if invoked directly.
+  await validateProfileForProvision(profile, opts);
+
+  const { seats, yaml } = await generateRoster(profile, opts);
+  const summary = formatTopologySummary(seats);
+
+  if (!opts.write) {
+    return { yaml, summary };
+  }
+
+  const rosterPath = join(dirs.mosaicHome, 'fleet', 'roster.yaml');
+  if (!opts.force) {
+    let exists = false;
+    try {
+      await access(rosterPath, constants.F_OK);
+      exists = true;
+    } catch {
+      exists = false;
+    }
+    if (exists) {
+      throw new Error(
+        `Refusing to overwrite existing roster: ${rosterPath}. ` +
+          `Pass --force to overwrite, or edit it by hand.`,
+      );
+    }
+  }
+  await mkdir(dirname(rosterPath), { recursive: true });
+  await writeFile(rosterPath, yaml, 'utf8');
+  return { yaml, summary, wrote: rosterPath };
+}
+
+/**
+ * Register `provision` under an existing `fleet` command. `mosaicHomeFor`
+ * resolves the active --mosaic-home (parent flag) at call time, exactly like the
+ * profile/persona/backlog subcommands.
+ */
+export function registerFleetProvisionCommand(
+  fleetCmd: Command,
+  mosaicHomeFor: () => string,
+): Command {
+  const provisionCmd = fleetCmd
+    .command('provision')
+    .description('Materialize a roster.yaml from a system-type profile (H3). DRY-RUN by default.')
+    .requiredOption('--profile <id>', 'System-type profile id to provision')
+    .option('--full', 'Materialize the entire profile roster (default: floor seats only)')
+    .option('--write', 'Write the generated roster to <mosaicHome>/fleet/roster.yaml')
+    .option('--force', 'Overwrite an existing roster.yaml (requires --write)')
+    .action(async (opts: { profile: string; full?: boolean; write?: boolean; force?: boolean }) => {
+      try {
+        const result = await runProvision(opts.profile, {
+          mosaicHome: mosaicHomeFor(),
+          full: opts.full,
+          write: opts.write,
+          force: opts.force,
+        });
+        if (result.wrote) {
+          console.log(`Wrote roster: ${result.wrote}`);
+          console.log('');
+          console.log(result.summary);
+        } else {
+          // DRY RUN: print the roster it WOULD generate + topology, write nothing.
+          console.log('# DRY RUN — no files written. Re-run with --write to persist.');
+          console.log(result.yaml.trimEnd());
+          console.log('');
+          console.log(result.summary);
+        }
+      } catch (err) {
+        process.stderr.write(`${err instanceof Error ? err.message : String(err)}\n`);
+        process.exitCode = 1;
+      }
+    });
+
+  return provisionCmd;
+}

packages/mosaic/src/commands/fleet.spec.ts

@@ -4,6 +4,7 @@ import { dirname, join, resolve } from 'node:path';
 import { Command } from 'commander';
 import { afterEach, describe, expect, it, vi } from 'vitest';
 import {
+  acquireRestartLock,
   addAgentToRoster,
   buildAgentSendCommand,
   buildAgentWatchAttachCommand,
@@ -45,6 +46,8 @@ import {
   removeAgentFromRoster,
   resolveFleetPaths,
   resolvePresetFilename,
+  restartLockPath,
+  RESTART_LOCK_STALE_MS,
   RUNTIME_ACCEPTABLE_COMMANDS,
   serializeRosterToYaml,
   VERIFY_DEFAULT_TIMEOUT_MS,
@@ -84,6 +87,7 @@ describe('registerFleetCommand', () => {
       'install-systemd',
       'persona',
       'profile',
+      'provision',
       'ps',
       'remove',
       'restart',
@@ -677,6 +681,364 @@ describe('fleet command construction', () => {
     }
   });
 
+  it('waits for an in-flight restart to clear before relaunching (re-entry guard)', async () => {
+    const home = await tempDir();
+    const rosterPath = join(home, 'fleet', 'roster.yaml');
+    await mkdir(join(home, 'fleet'), { recursive: true });
+    await writeFile(
+      rosterPath,
+      ['version: 1', 'transport: tmux', 'agents:', '  - name: coder0', '    runtime: codex'].join(
+        '\n',
+      ),
+    );
+
+    // Simulate another `mosaic fleet restart` process mid-teardown: a fresh lock
+    // (recent timestamp, so it is NOT treated as stale) already held.
+    const lockPath = restartLockPath(home);
+    await mkdir(dirname(lockPath), { recursive: true });
+    await writeFile(lockPath, `4242\n${Date.now()}\n`);
+
+    const events: string[] = [];
+    const runner: CommandRunner = async (command, args) => {
+      events.push(`run:${args[args.length - 1]}`);
+      return { stdout: '', stderr: '', exitCode: 0 };
+    };
+    // The injected sleep stands in for time passing while we wait; the in-flight
+    // restart "finishes" (releases its lock) after the first poll.
+    let sleeps = 0;
+    const sleepFn: SleepFn = async () => {
+      sleeps += 1;
+      events.push(`sleep:${sleeps}`);
+      await rm(lockPath, { force: true });
+    };
+
+    const program = new Command();
+    program.exitOverride();
+    registerFleetCommand(program, { runner, sleepFn, mosaicHome: home });
+
+    try {
+      await program.parseAsync(['node', 'mosaic', 'fleet', 'restart']);
+
+      // It must have waited at least once before issuing any systemctl restart.
+      expect(sleeps).toBeGreaterThan(0);
+      const firstSleep = events.findIndex((e) => e.startsWith('sleep:'));
+      const firstRun = events.findIndex((e) => e.startsWith('run:'));
+      expect(firstSleep).toBeGreaterThanOrEqual(0);
+      expect(firstRun).toBeGreaterThan(firstSleep);
+
+      // And it still performs the full restart once the lock clears.
+      expect(events).toContain('run:mosaic-tmux-holder.service');
+      expect(events).toContain('run:mosaic-agent@coder0.service');
+
+      // The lock is released after the restart completes.
+      await expect(readFile(lockPath, 'utf8')).rejects.toMatchObject({ code: 'ENOENT' });
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('breaks a stale restart lock and proceeds without waiting', async () => {
+    const home = await tempDir();
+    const rosterPath = join(home, 'fleet', 'roster.yaml');
+    await mkdir(join(home, 'fleet'), { recursive: true });
+    await writeFile(
+      rosterPath,
+      ['version: 1', 'transport: tmux', 'agents:', '  - name: coder0', '    runtime: codex'].join(
+        '\n',
+      ),
+    );
+
+    // A lock left behind by a crashed owner: timestamp older than the stale window.
+    const lockPath = restartLockPath(home);
+    await mkdir(dirname(lockPath), { recursive: true });
+    await writeFile(lockPath, `4242\n${Date.now() - RESTART_LOCK_STALE_MS - 1_000}\n`);
+
+    const calls: string[][] = [];
+    const runner: CommandRunner = async (command, args) => {
+      calls.push([command, ...args]);
+      return { stdout: '', stderr: '', exitCode: 0 };
+    };
+    const sleepFn = vi.fn<SleepFn>(async () => {});
+
+    const program = new Command();
+    program.exitOverride();
+    registerFleetCommand(program, { runner, sleepFn, mosaicHome: home });
+
+    try {
+      await program.parseAsync(['node', 'mosaic', 'fleet', 'restart']);
+
+      // Stale lock is broken immediately — no waiting.
+      expect(sleepFn).not.toHaveBeenCalled();
+      expect(calls).toEqual([
+        ['systemctl', '--user', 'restart', 'mosaic-tmux-holder.service'],
+        ['systemctl', '--user', 'restart', 'mosaic-agent@coder0.service'],
+      ]);
+      // The stale lock is gone once the restart completes.
+      await expect(readFile(lockPath, 'utf8')).rejects.toMatchObject({ code: 'ENOENT' });
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('releases the restart lock so a subsequent restart is not blocked', async () => {
+    const home = await tempDir();
+    const rosterPath = join(home, 'fleet', 'roster.yaml');
+    await mkdir(join(home, 'fleet'), { recursive: true });
+    await writeFile(
+      rosterPath,
+      ['version: 1', 'transport: tmux', 'agents:', '  - name: coder0', '    runtime: codex'].join(
+        '\n',
+      ),
+    );
+
+    const calls: string[][] = [];
+    const runner: CommandRunner = async (command, args) => {
+      calls.push([command, ...args]);
+      return { stdout: '', stderr: '', exitCode: 0 };
+    };
+    const sleepFn = vi.fn<SleepFn>(async () => {});
+
+    const program = new Command();
+    program.exitOverride();
+    registerFleetCommand(program, { runner, sleepFn, mosaicHome: home });
+
+    try {
+      await program.parseAsync(['node', 'mosaic', 'fleet', 'restart']);
+      await program.parseAsync(['node', 'mosaic', 'fleet', 'restart']);
+
+      // Two sequential restarts both run fully and neither has to wait.
+      expect(sleepFn).not.toHaveBeenCalled();
+      expect(calls).toEqual([
+        ['systemctl', '--user', 'restart', 'mosaic-tmux-holder.service'],
+        ['systemctl', '--user', 'restart', 'mosaic-agent@coder0.service'],
+        ['systemctl', '--user', 'restart', 'mosaic-tmux-holder.service'],
+        ['systemctl', '--user', 'restart', 'mosaic-agent@coder0.service'],
+      ]);
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('guards the single-agent restart path behind the in-flight restart lock', async () => {
+    const home = await tempDir();
+    const rosterPath = join(home, 'fleet', 'roster.yaml');
+    await mkdir(join(home, 'fleet'), { recursive: true });
+    await writeFile(
+      rosterPath,
+      ['version: 1', 'transport: tmux', 'agents:', '  - name: coder0', '    runtime: codex'].join(
+        '\n',
+      ),
+    );
+
+    // A full restart is mid-flight (lock held); a single-agent restart re-enters.
+    const lockPath = restartLockPath(home);
+    await mkdir(dirname(lockPath), { recursive: true });
+    await writeFile(lockPath, `4242\n${Date.now()}\n`);
+
+    const events: string[] = [];
+    const runner: CommandRunner = async (command, args) => {
+      events.push(`run:${args[args.length - 1]}`);
+      return { stdout: '', stderr: '', exitCode: 0 };
+    };
+    let sleeps = 0;
+    const sleepFn: SleepFn = async () => {
+      sleeps += 1;
+      events.push(`sleep:${sleeps}`);
+      await rm(lockPath, { force: true });
+    };
+
+    const program = new Command();
+    program.exitOverride();
+    registerFleetCommand(program, { runner, sleepFn, mosaicHome: home });
+
+    try {
+      await program.parseAsync(['node', 'mosaic', 'fleet', 'restart', 'coder0']);
+
+      // The single-agent restart waits for the in-flight restart before acting.
+      expect(sleeps).toBeGreaterThan(0);
+      const firstSleep = events.findIndex((e) => e.startsWith('sleep:'));
+      const firstRun = events.findIndex((e) => e.startsWith('run:'));
+      expect(firstSleep).toBeGreaterThanOrEqual(0);
+      expect(firstRun).toBeGreaterThan(firstSleep);
+      // Only the named agent is restarted; the holder is untouched.
+      expect(events).toContain('run:mosaic-agent@coder0.service');
+      expect(events).not.toContain('run:mosaic-tmux-holder.service');
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('does not let a timed-out owner drop a lock another restart broke and re-owned', async () => {
+    const home = await tempDir();
+    const runDir = join(home, 'fleet', 'run');
+    await mkdir(runDir, { recursive: true });
+    const lockPath = restartLockPath(home);
+    const tokenOf = async (): Promise<string> => {
+      const raw = await readFile(lockPath, 'utf8');
+      return raw.split('\n')[2]?.trim() ?? '';
+    };
+    const sleepFn = vi.fn<SleepFn>(async () => {});
+
+    // R1 acquires the lock and begins a restart that then hangs.
+    const r1 = await acquireRestartLock(home, sleepFn);
+    const tokenR1 = await tokenOf();
+    expect(tokenR1).not.toBe('');
+
+    // The hung R1 leaves a stale lock: rewrite its timestamp into the past while
+    // preserving R1's token — exactly the on-disk state a stuck owner leaves.
+    await writeFile(lockPath, `4242\n${Date.now() - RESTART_LOCK_STALE_MS - 1_000}\n${tokenR1}\n`);
+
+    // R2 re-enters, sees the stale lock, and atomically takes ownership.
+    const r2 = await acquireRestartLock(home, sleepFn);
+    const tokenR2 = await tokenOf();
+    expect(tokenR2).not.toBe(tokenR1);
+    expect(sleepFn).not.toHaveBeenCalled();
+
+    // R1 finally finishes and releases. It must NOT delete R2's lock — otherwise
+    // a third restart (R3) could acquire and interleave with R2 still running.
+    await r1.release();
+    expect(await tokenOf()).toBe(tokenR2);
+
+    // R2 releases cleanly and the lock is gone.
+    await r2.release();
+    await expect(readFile(lockPath, 'utf8')).rejects.toMatchObject({ code: 'ENOENT' });
+
+    await rm(home, { recursive: true, force: true });
+  });
+
+  it('lets only one of several concurrent breakers proceed past a stale lock', async () => {
+    const home = await tempDir();
+    const lockPath = restartLockPath(home);
+    await mkdir(dirname(lockPath), { recursive: true });
+
+    // A stale lock left by a crashed owner: every concurrent re-entrant restart
+    // will judge it stale and try to break it at the same instant. Breaking must
+    // NOT grant ownership — only the atomic re-create may — so exactly one
+    // contender can ever hold the lock at a time. (The v2 fix wrote our own token
+    // during the break and read it back, so two breakers each saw their own token
+    // and BOTH proceeded; this guards that regression.)
+    await writeFile(
+      lockPath,
+      `4242\n${Date.now() - RESTART_LOCK_STALE_MS - 1_000}\nstale-owner-token\n`,
+    );
+
+    // Yielding sleep so a waiting contender lets the current owner finish and
+    // release before it re-contends, instead of spinning the microtask queue.
+    const sleepFn: SleepFn = async () => {
+      await new Promise((res) => setTimeout(res, 0));
+    };
+
+    let active = 0;
+    let maxActive = 0;
+    const tokens: string[] = [];
+    const tokenOf = async (): Promise<string> => {
+      const raw = await readFile(lockPath, 'utf8');
+      return raw.split('\n')[2]?.trim() ?? '';
+    };
+
+    // One "restart" = acquire the lock, do work in the critical section, release.
+    const restartOnce = async (): Promise<void> => {
+      const guard = await acquireRestartLock(home, sleepFn);
+      active += 1;
+      maxActive = Math.max(maxActive, active);
+      // Record the token we own while we hold it, then yield to interleave with
+      // any other contender that might (wrongly) believe it owns the lock too.
+      tokens.push(await tokenOf());
+      await new Promise((res) => setTimeout(res, 0));
+      active -= 1;
+      await guard.release();
+    };
+
+    try {
+      // Three breakers race the single stale lock simultaneously.
+      await Promise.all([restartOnce(), restartOnce(), restartOnce()]);
+
+      // Mutual exclusion held: never two owners at once despite concurrent breaks.
+      expect(maxActive).toBe(1);
+      // Each acquire owned with its own distinct token — no two ever shared it.
+      expect(new Set(tokens).size).toBe(3);
+      // The lock is fully released at the end.
+      await expect(readFile(lockPath, 'utf8')).rejects.toMatchObject({ code: 'ENOENT' });
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
+  it('lets exactly one of two breakers take over a stale lock while the other waits', async () => {
+    const home = await tempDir();
+    const lockPath = restartLockPath(home);
+    await mkdir(dirname(lockPath), { recursive: true });
+
+    // A single stale lock both contenders will judge stale at the same instant.
+    // Every transition runs under the registry mutex, so only one may take the
+    // lock over; the other must observe a now-fresh owner and WAIT/re-evaluate
+    // rather than also taking over. (A content-blind clobber let both believe
+    // they owned it — this asserts the mutex-gated CAS takeover instead.)
+    await writeFile(
+      lockPath,
+      `4242\n${Date.now() - RESTART_LOCK_STALE_MS - 1_000}\nstale-owner-token\n`,
+    );
+
+    // Barrier the winner holds against until the loser has observed the lock
+    // fresh and waited at least once — forcing the exact interleaving where one
+    // proceeds while the other waits, deterministically rather than by timing.
+    let resolveLoserWaited: () => void = () => {};
+    const loserWaited = new Promise<void>((res) => {
+      resolveLoserWaited = res;
+    });
+    let sleeps = 0;
+    const sleepFn: SleepFn = async () => {
+      sleeps += 1;
+      resolveLoserWaited();
+      await new Promise((res) => setTimeout(res, 0));
+    };
+
+    let active = 0;
+    let maxActive = 0;
+    const tokens: string[] = [];
+    const tokenOf = async (): Promise<string> => {
+      const raw = await readFile(lockPath, 'utf8');
+      return raw.split('\n')[2]?.trim() ?? '';
+    };
+
+    let firstOwner = true;
+    const restartOnce = async (): Promise<void> => {
+      const guard = await acquireRestartLock(home, sleepFn);
+      active += 1;
+      maxActive = Math.max(maxActive, active);
+      tokens.push(await tokenOf());
+      if (firstOwner) {
+        // Winner: keep holding the lock until the loser has waited once, so the
+        // loser is guaranteed to see a FRESH owner (not the stale one) and back
+        // off — proving it could not also take over.
+        firstOwner = false;
+        await loserWaited;
+      } else {
+        await new Promise((res) => setTimeout(res, 0));
+      }
+      active -= 1;
+      await guard.release();
+    };
+
+    try {
+      // Exactly two breakers race the single stale lock.
+      await Promise.all([restartOnce(), restartOnce()]);
+
+      // Mutual exclusion: never two owners at once (if both took over the stale
+      // lock, this would be 2).
+      expect(maxActive).toBe(1);
+      // Both eventually owned, each with its own distinct token.
+      expect(new Set(tokens).size).toBe(2);
+      // The loser observed the winner's fresh lock and waited — it did NOT also
+      // take over the stale lock.
+      expect(sleeps).toBeGreaterThanOrEqual(1);
+      // The lock is fully released at the end.
+      await expect(readFile(lockPath, 'utf8')).rejects.toMatchObject({ code: 'ENOENT' });
+    } finally {
+      await rm(home, { recursive: true, force: true });
+    }
+  });
+
   it('attempts every agent and the holder during fleet stop even when an agent stop fails', async () => {
     const home = await tempDir();
     const rosterPath = join(home, 'fleet', 'roster.yaml');

packages/mosaic/src/commands/fleet.ts

@@ -1,5 +1,16 @@
 import { constants } from 'node:fs';
-import { access, chmod, copyFile, mkdir, readFile, unlink, writeFile } from 'node:fs/promises';
+import {
+  access,
+  chmod,
+  copyFile,
+  mkdir,
+  open,
+  readFile,
+  stat,
+  unlink,
+  writeFile,
+} from 'node:fs/promises';
+import { randomUUID } from 'node:crypto';
 import { homedir, hostname, userInfo } from 'node:os';
 import { dirname, join, resolve } from 'node:path';
 import { fileURLToPath } from 'node:url';
@@ -11,6 +22,7 @@ import { resolveCommsBlock } from '../fleet/comms-onboarding.js';
 import { registerFleetBacklogCommand } from './fleet-backlog.js';
 import { registerFleetPersonaCommand } from './fleet-personas.js';
 import { registerFleetProfileCommand } from './fleet-profiles.js';
+import { registerFleetProvisionCommand } from './fleet-provision.js';
 
 /**
  * A function that spawns a command with inherited stdio (TTY passthrough).
@@ -532,6 +544,295 @@ export function buildFleetServiceCommand(action: FleetServiceAction, agentName?:
   return ['systemctl', '--user', action, service];
 }
 
+/** Poll interval (ms) while waiting for an in-flight restart's lock to clear. */
+export const RESTART_LOCK_POLL_INTERVAL_MS = 250;
+/**
+ * Maximum time (ms) a re-entrant restart waits for the in-flight restart to
+ * finish before it breaks the lock and proceeds anyway. A bound is required so
+ * a crashed holder of the lock can never deadlock the fleet permanently.
+ */
+export const RESTART_LOCK_MAX_WAIT_MS = 30_000;
+/**
+ * Age (ms) past which a restart lock is treated as stale (its owner died
+ * without releasing it) and is broken immediately rather than waited on.
+ */
+export const RESTART_LOCK_STALE_MS = 60_000;
+
+/**
+ * Resolves the path of the cross-process restart lock for a given Mosaic home.
+ * Kept strictly under `<mosaicHome>/fleet/run` (not the heartbeat env override)
+ * so the lock is scoped to the same fleet the restart acts on.
+ */
+export function restartLockPath(mosaicHome: string): string {
+  return join(mosaicHome, 'fleet', 'run', 'restart.lock');
+}
+
+/** A held restart lock; `release()` removes the lock file iff we still own it. */
+interface RestartGuard {
+  release(): Promise<void>;
+}
+
+/** Lock-file contents: pid (informational), timestamp, and a unique owner token. */
+function formatRestartLockContent(token: string): string {
+  return `${process.pid}\n${Date.now()}\n${token}\n`;
+}
+
+/**
+ * Reads the owner token (line 3) from a lock file, or null if the file is
+ * missing/unreadable/tokenless. The token is what makes release and break
+ * ownership-safe: a process only ever acts on a lock whose token matches its own.
+ */
+async function readRestartLockToken(lockPath: string): Promise<string | null> {
+  let raw: string;
+  try {
+    raw = await readFile(lockPath, 'utf8');
+  } catch {
+    return null;
+  }
+  const token = raw.split('\n')[2]?.trim();
+  return token ? token : null;
+}
+
+/**
+ * Returns true when a lock's contents are stale: older than RESTART_LOCK_STALE_MS,
+ * or unparseable (a corrupt or partially written lock left by a crashed owner).
+ */
+function isRestartLockContentStale(raw: string, now: number): boolean {
+  const stampLine = raw.split('\n')[1] ?? '';
+  const stamp = Number.parseInt(stampLine.trim(), 10);
+  if (!Number.isFinite(stamp)) {
+    return true;
+  }
+  return now - stamp >= RESTART_LOCK_STALE_MS;
+}
+
+/**
+ * Path of the short-lived registry mutex that guards EVERY transition of the
+ * restart lock (acquire, release, takeover). Held only across a few filesystem
+ * ops — never across the restart itself — so contention clears in microseconds.
+ */
+function restartMutexPath(lockPath: string): string {
+  return `${lockPath}.mutex`;
+}
+
+/** Brief back-off between registry-mutex acquisition attempts (held microseconds). */
+const RESTART_MUTEX_RETRY_MS = 20;
+
+/**
+ * Staleness for the internal mutex / reclaim locks, judged by the file's mtime
+ * rather than its CONTENT. `open(path, 'wx')` creates the inode (with a fresh
+ * mtime) before any token/timestamp is written into it, so a content-based check
+ * would momentarily see that empty file as corrupt-and-stale and could reap a
+ * lock another contender is still acquiring. mtime is set atomically at creation,
+ * so a just-created lock always reads as live; only a lock whose holder died and
+ * stopped touching it ages past the threshold. These locks are never held across
+ * the restart itself (only a couple of filesystem ops), so any mtime this old can
+ * belong only to a dead holder.
+ */
+async function isRestartLockPathStale(path: string, now: number): Promise<boolean> {
+  try {
+    const info = await stat(path);
+    return now - info.mtimeMs >= RESTART_LOCK_STALE_MS;
+  } catch (err) {
+    if ((err as NodeJS.ErrnoException).code === 'ENOENT') {
+      return false; // Gone, not stale — the caller will re-contend.
+    }
+    return false; // Can't stat — treat as live and back off rather than reap.
+  }
+}
+
+/** Path of the reclaim lock that serializes reaping of a crashed-holder mutex. */
+function restartReclaimPath(mutexPath: string): string {
+  return `${mutexPath}.reclaim`;
+}
+
+/**
+ * Reap a registry mutex left behind by a process that CRASHED mid-transition —
+ * one whose file has aged past RESTART_LOCK_STALE_MS. Because the mutex is held
+ * only for a couple of filesystem ops (no sleeps, never across the restart), a
+ * mutex this old can only belong to a dead holder.
+ *
+ * The reap removes the dead mutex but never CREATES/holds it — acquisition stays
+ * the single `open('wx')` create in {@link acquireRestartMutex}, so exactly one
+ * contender wins ownership no matter how the reap and acquires interleave. The
+ * removal is made conditional by a dedicated reclaim lock: while it is held the
+ * dead mutex is stable (its dead holder will never touch it, and no other
+ * reclaimer can race), so re-reading it and removing it only if it is STILL stale
+ * is a true compare — a live holder's fresh mutex is never removed. This closes
+ * the reclaim race a content-blind rename-and-restore left open (a third
+ * contender slipping into the gap while a fresh mutex was moved aside).
+ */
+async function reclaimStaleRestartMutex(mutexPath: string): Promise<void> {
+  const reclaimPath = restartReclaimPath(mutexPath);
+  let handle: Awaited<ReturnType<typeof open>>;
+  try {
+    handle = await open(reclaimPath, 'wx');
+  } catch (err) {
+    if ((err as NodeJS.ErrnoException).code !== 'EEXIST') {
+      throw err;
+    }
+    // Someone is already reclaiming. If their reclaim lock is itself stale by
+    // mtime, its holder crashed mid-reap (the lock spans only a stat + unlink,
+    // microseconds) — clear it so a later pass can retry. Otherwise a live
+    // reclaimer has it; back off. Either way we do not reap the mutex this pass.
+    if (await isRestartLockPathStale(reclaimPath, Date.now())) {
+      await unlink(reclaimPath).catch(() => {});
+    }
+    return;
+  }
+  try {
+    // Re-check the mutex UNDER the reclaim lock and remove it only if it is STILL
+    // stale by mtime. A live holder's mutex is fresh and is left untouched; a dead
+    // holder's mutex is stable here (its holder is gone and no other reclaimer can
+    // race us), so this re-check is authoritative.
+    if (await isRestartLockPathStale(mutexPath, Date.now())) {
+      await unlink(mutexPath).catch(() => {});
+    }
+  } finally {
+    await handle.close();
+    await unlink(reclaimPath).catch(() => {});
+  }
+}
+
+/**
+ * Acquire the registry mutex, BLOCKING (with brief back-offs) until held, and
+ * return a token-gated release. This is the single point of mutual exclusion for
+ * the restart lock: acquire, release, and stale/timeout takeover all run under it,
+ * so "read the lock, then mutate it" is atomic — no acquirer, releaser, or breaker
+ * can ever interleave with another. A mutex left by a crashed holder is reclaimed
+ * once it ages past the stale threshold.
+ */
+async function acquireRestartMutex(
+  mutexPath: string,
+  token: string,
+): Promise<RestartGuard['release']> {
+  for (;;) {
+    let handle: Awaited<ReturnType<typeof open>>;
+    try {
+      handle = await open(mutexPath, 'wx');
+    } catch (err) {
+      if ((err as NodeJS.ErrnoException).code !== 'EEXIST') {
+        throw err;
+      }
+      // Staleness is judged by mtime, not content, so a mutex that exists but has
+      // not yet had its token written (the open-before-write window) reads as live
+      // and is never wrongly reaped.
+      if (!(await isRestartLockPathStale(mutexPath, Date.now()))) {
+        // A live holder has it — it will be gone in microseconds. Back off briefly.
+        await new Promise((resolve) => setTimeout(resolve, RESTART_MUTEX_RETRY_MS));
+        continue;
+      }
+      await reclaimStaleRestartMutex(mutexPath);
+      continue;
+    }
+    // We created the mutex. Populate it with our token; if writing fails, clean up
+    // our own file so we never leak an empty mutex that a peer would have to reap.
+    try {
+      await handle.writeFile(formatRestartLockContent(token));
+      await handle.close();
+    } catch (err) {
+      await handle.close().catch(() => {});
+      await unlink(mutexPath).catch(() => {});
+      throw err;
+    }
+    return async (): Promise<void> => {
+      if ((await readRestartLockToken(mutexPath)) !== token) return;
+      await unlink(mutexPath).catch(() => {});
+    };
+  }
+}
+
+/**
+ * Acquire the fleet restart lock, serializing concurrent `mosaic fleet restart`
+ * invocations across processes. Each restart tears the tmux holder (and the
+ * agent sessions inside it) down and back up; without this guard a re-entrant
+ * restart relaunches agents against a half-torn-down holder, which fails and
+ * tight-loops. A re-entrant caller waits for the in-flight restart to release
+ * the lock (clean shutdown settled) before proceeding, breaks a stale lock left
+ * by a crashed owner, and after RESTART_LOCK_MAX_WAIT_MS breaks the lock to
+ * avoid a permanent deadlock.
+ *
+ * Correctness rests on a single invariant: EVERY transition of the lock — taking
+ * a free lock, taking over a stale/timed-out one, and releasing — happens under
+ * the registry mutex. Because the check ("is the lock free / stale / fresh?") and
+ * the mutation that follows it both run while the mutex is held, they are atomic:
+ * no other acquirer, releaser, or breaker can slip in between. That is what makes
+ * takeover a true compare-and-swap rather than a content-blind clobber — a normal
+ * `open('wx')` acquirer cannot create a fresh lock in a gap, and the original
+ * owner's `release()` (also mutex-gated and token-checked) cannot drop a lock a
+ * breaker already took over. So no interleaving lets two restarts both own the
+ * lock and run concurrently.
+ */
+export async function acquireRestartLock(
+  mosaicHome: string,
+  sleepFn: SleepFn,
+): Promise<RestartGuard> {
+  const token = randomUUID();
+  const lockPath = restartLockPath(mosaicHome);
+  const mutexPath = restartMutexPath(lockPath);
+  await mkdir(dirname(lockPath), { recursive: true });
+  const release = async (): Promise<void> => {
+    // Mutex-gated and token-gated: only remove the lock if it is still ours. If
+    // another caller took it over (after a stale/timeout break) the token no
+    // longer matches and we leave their lock intact.
+    const releaseMutex = await acquireRestartMutex(mutexPath, token);
+    try {
+      if ((await readRestartLockToken(lockPath)) === token) {
+        await unlink(lockPath).catch(() => {});
+      }
+    } finally {
+      await releaseMutex();
+    }
+  };
+  const deadline = Date.now() + RESTART_LOCK_MAX_WAIT_MS;
+  for (;;) {
+    let owned = false;
+    const releaseMutex = await acquireRestartMutex(mutexPath, token);
+    try {
+      // Read and (if appropriate) mutate the lock atomically under the mutex.
+      let current: string | null = null;
+      let absent = false;
+      try {
+        current = await readFile(lockPath, 'utf8');
+      } catch (readErr) {
+        if ((readErr as NodeJS.ErrnoException).code === 'ENOENT') {
+          absent = true;
+        } else {
+          current = null; // Unreadable/corrupt: treat as stale.
+        }
+      }
+      const now = Date.now();
+      if (absent) {
+        // Lock is free — take it.
+        await writeFile(lockPath, formatRestartLockContent(token));
+        owned = true;
+      } else {
+        const stale = current === null || isRestartLockContentStale(current, now);
+        const timedOut = now >= deadline;
+        if (stale || timedOut) {
+          process.stderr.write(
+            stale
+              ? 'Breaking stale fleet restart lock.\n'
+              : `Timed out after ${RESTART_LOCK_MAX_WAIT_MS}ms waiting for the in-flight fleet ` +
+                  'restart; breaking the lock.\n',
+          );
+          // Takeover is just an overwrite — safe because we hold the mutex, so no
+          // acquirer or releaser can touch the lock between our read and this write.
+          await writeFile(lockPath, formatRestartLockContent(token));
+          owned = true;
+        }
+        // else: a fresh restart owns it — wait below and re-evaluate.
+      }
+    } finally {
+      await releaseMutex();
+    }
+    if (owned) {
+      return { release };
+    }
+    await sleepFn(RESTART_LOCK_POLL_INTERVAL_MS);
+  }
+}
+
 /**
  * Returns the systemctl --user enable command for a given unit.
  * Used by the install auto-enable step to persist units across reboots.
@@ -1171,6 +1472,7 @@ export function isSendAccepted(capturedOutput: string): SendVerifyResult {
 
 export function registerFleetCommand(program: Command, deps: FleetCommandDeps = {}): Command {
   const runner = deps.runner ?? runCommand;
+  const sleepFn = deps.sleepFn ?? defaultSleep;
   const paths = resolveFleetPaths(deps.mosaicHome);
   const frameworkRoot = deps.frameworkRoot ?? resolveFrameworkRoot();
 
@@ -1284,9 +1586,22 @@ export function registerFleetCommand(program: Command, deps: FleetCommandDeps =
       .command(`${action} [agent]`)
       .description(`${action} the fleet holder or one agent`)
       .action(async (agent?: string) => {
+        const commandOpts = cmd.opts<{ mosaicHome: string; roster?: string }>();
+        const activePaths = resolveFleetPaths(commandOpts.mosaicHome);
         const roster = await loadRosterForCommand(cmd);
         if (agent) {
           getRosterAgent(roster, agent);
+          // Single-agent restart is guarded too: it can race a full restart that
+          // is tearing the shared holder down.
+          if (action === 'restart') {
+            const guard = await acquireRestartLock(activePaths.mosaicHome, sleepFn);
+            try {
+              await runChecked(runner, buildFleetServiceCommand(action, agent));
+            } finally {
+              await guard.release();
+            }
+            return;
+          }
           await runChecked(runner, buildFleetServiceCommand(action, agent));
           return;
         }
@@ -1297,6 +1612,21 @@ export function registerFleetCommand(program: Command, deps: FleetCommandDeps =
           );
           return;
         }
+        if (action === 'restart') {
+          // Serialize the holder+agents teardown/relaunch behind the restart lock
+          // so a re-entrant restart waits for clean shutdown before relaunching,
+          // instead of racing a half-torn-down holder into a tight loop.
+          const guard = await acquireRestartLock(activePaths.mosaicHome, sleepFn);
+          try {
+            await runChecked(runner, buildFleetServiceCommand(action));
+            for (const rosterAgent of roster.agents) {
+              await runChecked(runner, buildFleetServiceCommand(action, rosterAgent.name));
+            }
+          } finally {
+            await guard.release();
+          }
+          return;
+        }
         await runChecked(runner, buildFleetServiceCommand(action));
         for (const rosterAgent of roster.agents) {
           await runChecked(runner, buildFleetServiceCommand(action, rosterAgent.name));
@@ -1720,6 +2050,10 @@ export function registerFleetCommand(program: Command, deps: FleetCommandDeps =
   // --mosaic-home flag.
   registerFleetPersonaCommand(cmd, () => cmd.opts<{ mosaicHome: string }>().mosaicHome);
 
+  // Provisioning (H3): materialize a concrete roster.yaml from a system-type
+  // profile. DRY-RUN by default; --write persists under the same --mosaic-home.
+  registerFleetProvisionCommand(cmd, () => cmd.opts<{ mosaicHome: string }>().mosaicHome);
+
   return cmd;
 }
 

packages/mosaic/vitest.config.ts

@@ -4,5 +4,6 @@ export default defineConfig({
   test: {
     globals: true,
     environment: 'node',
+    testTimeout: 30_000,
   },
 });

plugins/discord/src/index.ts

@@ -1,24 +1,109 @@
+import { createHmac, randomUUID, timingSafeEqual } from 'node:crypto';
 import { ChannelType, Client, GatewayIntentBits, type Message as DiscordMessage } from 'discord.js';
 import { io, type Socket } from 'socket.io-client';
 
 export interface DiscordPluginConfig {
   token: string;
   gatewayUrl: string;
-  /** Which guild to bind to (single-guild only for v0.1.0) */
+  /** Shared service credential injected by the approved secret mechanism. */
+  serviceToken: string;
+  /** Which guild to bind to (single-guild only for v0.1.0). */
   guildId?: string;
+  allowedGuildIds: readonly string[];
+  allowedChannelIds: readonly string[];
+  allowedUserIds: readonly string[];
+}
+
+export interface DiscordIngressPayload {
+  correlationId: string;
+  messageId: string;
+  guildId: string;
+  channelId: string;
+  userId: string;
+  conversationId: string;
+  content: string;
+}
+
+export interface DiscordIngressEnvelope {
+  payload: DiscordIngressPayload;
+  signature: string;
+}
+
+export interface DiscordIngressAllowlists {
+  guildIds: readonly string[];
+  channelIds: readonly string[];
+  userIds: readonly string[];
+}
+
+function signedPayload(payload: DiscordIngressPayload): string {
+  return [
+    payload.correlationId,
+    payload.messageId,
+    payload.guildId,
+    payload.channelId,
+    payload.userId,
+    payload.conversationId,
+    payload.content,
+  ].join('\n');
+}
+
+function signPayload(payload: DiscordIngressPayload, serviceToken: string): string {
+  return createHmac('sha256', serviceToken).update(signedPayload(payload)).digest('hex');
+}
+
+function isSignatureValid(actual: string, expected: string): boolean {
+  const actualBuffer = Buffer.from(actual, 'hex');
+  const expectedBuffer = Buffer.from(expected, 'hex');
+  return (
+    actualBuffer.length === expectedBuffer.length && timingSafeEqual(actualBuffer, expectedBuffer)
+  );
+}
+
+function includesId(allowedIds: readonly string[], id: string): boolean {
+  return allowedIds.includes(id);
+}
+
+/** Creates the signed, auditable envelope accepted by the gateway Discord service boundary. */
+export function createDiscordIngressEnvelope(
+  payload: DiscordIngressPayload,
+  serviceToken: string,
+): DiscordIngressEnvelope {
+  return { payload, signature: signPayload(payload, serviceToken) };
+}
+
+/**
+ * Verifies service-origin integrity and applies default-deny Discord identity allowlists.
+ * Returns null instead of a partially trusted payload on every failure path.
+ */
+export function verifyDiscordIngressEnvelope(
+  envelope: DiscordIngressEnvelope,
+  serviceToken: string,
+  allowlists?: DiscordIngressAllowlists,
+): DiscordIngressPayload | null {
+  const expectedSignature = signPayload(envelope.payload, serviceToken);
+  if (!isSignatureValid(envelope.signature, expectedSignature)) return null;
+
+  if (
+    allowlists &&
+    (!includesId(allowlists.guildIds, envelope.payload.guildId) ||
+      !includesId(allowlists.channelIds, envelope.payload.channelId) ||
+      !includesId(allowlists.userIds, envelope.payload.userId))
+  ) {
+    return null;
+  }
+
+  return envelope.payload;
 }
 
 export class DiscordPlugin {
   private client: Client;
   private socket: Socket | null = null;
-  private config: DiscordPluginConfig;
-  /** Map Discord channel ID → Mosaic conversation ID */
+  /** Map Discord channel ID → Mosaic conversation ID. */
   private channelConversations = new Map<string, string>();
-  /** Track in-flight responses to avoid duplicate streaming */
+  /** Track in-flight responses to avoid duplicate streaming. */
   private pendingResponses = new Map<string, string>();
 
-  constructor(config: DiscordPluginConfig) {
-    this.config = config;
+  constructor(private readonly config: DiscordPluginConfig) {
     this.client = new Client({
       intents: [
         GatewayIntentBits.Guilds,
@@ -30,8 +115,8 @@ export class DiscordPlugin {
   }
 
   async start(): Promise<void> {
-    // Connect to gateway WebSocket
     this.socket = io(`${this.config.gatewayUrl}/chat`, {
+      auth: { discordServiceToken: this.config.serviceToken },
       transports: ['websocket'],
     });
 
@@ -48,7 +133,6 @@ export class DiscordPlugin {
       console.error(`[discord] Gateway connection error: ${err.message}`);
     });
 
-    // Handle streaming text from gateway
     this.socket.on('agent:text', (data: { conversationId: string; text: string }) => {
       const pending = this.pendingResponses.get(data.conversationId);
       if (pending !== undefined) {
@@ -56,12 +140,11 @@ export class DiscordPlugin {
       }
     });
 
-    // When agent finishes, send the accumulated response
     this.socket.on('agent:end', (data: { conversationId: string }) => {
       const text = this.pendingResponses.get(data.conversationId);
       if (text) {
         this.pendingResponses.delete(data.conversationId);
-        this.sendToDiscord(data.conversationId, text).catch((err) => {
+        this.sendToDiscord(data.conversationId, text).catch((err: unknown) => {
           console.error(`[discord] Error sending response for ${data.conversationId}:`, err);
         });
       }
@@ -71,8 +154,9 @@ export class DiscordPlugin {
       this.pendingResponses.set(data.conversationId, '');
     });
 
-    // Set up Discord message handler
-    this.client.on('messageCreate', (message) => this.handleDiscordMessage(message));
+    this.client.on('messageCreate', (message: DiscordMessage) =>
+      this.handleDiscordMessage(message),
+    );
 
     this.client.on('ready', () => {
       console.log(`[discord] Bot logged in as ${this.client.user?.tag}`);
@@ -96,7 +180,6 @@ export class DiscordPlugin {
     const guild = this.client.guilds.cache.get(this.config.guildId);
     if (!guild) return null;
 
-    // Slugify project name for channel: lowercase, replace spaces/special chars with hyphens
     const channelName = `mosaic-${project.name
       .toLowerCase()
       .replace(/[^a-z0-9]+/g, '-')
@@ -108,58 +191,58 @@ export class DiscordPlugin {
       topic: project.description ?? `Mosaic project: ${project.name}`,
     });
 
-    // Register the channel mapping so messages route correctly
     this.channelConversations.set(channel.id, `discord-${channel.id}`);
-
     return { channelId: channel.id };
   }
 
   private handleDiscordMessage(message: DiscordMessage): void {
-    // Ignore bot messages
-    if (message.author.bot) return;
+    if (message.author.bot || !this.client.user) return;
+    if (!message.guildId || !this.isAllowedMessage(message)) return;
 
-    // Not ready yet
-    if (!this.client.user) return;
-
-    // Check guild binding
-    if (this.config.guildId && message.guildId !== this.config.guildId) return;
-
-    // Respond to DMs always, or mentions in channels
-    const isDM = !message.guildId;
     const isMention = message.mentions.has(this.client.user);
+    if (!isMention) return;
 
-    if (!isDM && !isMention) return;
-
-    // Strip bot mention from message content
     const content = message.content
       .replace(new RegExp(`<@!?${this.client.user.id}>`, 'g'), '')
       .trim();
-
     if (!content) return;
-
-    // Get or create conversation for this Discord channel
-    const channelId = message.channelId;
-    let conversationId = this.channelConversations.get(channelId);
-    if (!conversationId) {
-      conversationId = `discord-${channelId}`;
-      this.channelConversations.set(channelId, conversationId);
-    }
-
-    // Send to gateway
     if (!this.socket?.connected) {
       console.error(
-        `[discord] Cannot forward message: not connected to gateway. channel=${channelId}`,
+        `[discord] Cannot forward message: not connected to gateway. channel=${message.channelId} message=${message.id}`,
       );
       return;
     }
-    this.socket.emit('message', {
-      conversationId,
-      content,
-    });
+
+    const channelId = message.channelId;
+    const conversationId = this.channelConversations.get(channelId) ?? `discord-${channelId}`;
+    this.channelConversations.set(channelId, conversationId);
+
+    const envelope = createDiscordIngressEnvelope(
+      {
+        correlationId: randomUUID(),
+        messageId: message.id,
+        guildId: message.guildId,
+        channelId,
+        userId: message.author.id,
+        conversationId,
+        content,
+      },
+      this.config.serviceToken,
+    );
+    this.socket.emit('message', envelope);
+  }
+
+  private isAllowedMessage(message: DiscordMessage): boolean {
+    const guildId = message.guildId;
+    return (
+      guildId !== null &&
+      includesId(this.config.allowedGuildIds, guildId) &&
+      includesId(this.config.allowedChannelIds, message.channelId) &&
+      includesId(this.config.allowedUserIds, message.author.id)
+    );
   }
 
   private async sendToDiscord(conversationId: string, text: string): Promise<void> {
-    // Find the Discord channel for this conversation
     const channelId = Array.from(this.channelConversations.entries()).find(
       ([, convId]) => convId === conversationId,
     )?.[0];
@@ -177,12 +260,10 @@ export class DiscordPlugin {
       return;
     }
 
-    // Chunk responses for Discord's 2000-char limit
-    const chunks = this.chunkText(text, 1900);
-    for (const chunk of chunks) {
+    for (const chunk of this.chunkText(text, 1900)) {
       try {
         await (channel as { send: (content: string) => Promise<unknown> }).send(chunk);
-      } catch (err) {
+      } catch (err: unknown) {
         console.error(`[discord] Failed to send message to channel ${channelId}:`, err);
       }
     }
@@ -193,21 +274,16 @@ export class DiscordPlugin {
 
     const chunks: string[] = [];
     let remaining = text;
-
     while (remaining.length > 0) {
       if (remaining.length <= maxLength) {
         chunks.push(remaining);
         break;
       }
-
-      // Try to break at a newline
       let breakPoint = remaining.lastIndexOf('\n', maxLength);
       if (breakPoint <= 0) breakPoint = maxLength;
-
       chunks.push(remaining.slice(0, breakPoint));
       remaining = remaining.slice(breakPoint).trimStart();
     }
-
     return chunks;
   }
 }

tools/install.sh

@@ -16,6 +16,10 @@
 #   --framework       Install/upgrade framework only (skip npm CLI)
 #   --cli             Install/upgrade npm CLI only (skip framework)
 #   --ref <branch>    Git ref for framework archive (default: main)
+#   --dev             Build CLI + gateway FROM SOURCE at --ref instead of the
+#                     registry @latest. Zero registry writes — packs local
+#                     tarballs and installs them globally. Use to test a branch
+#                     end-to-end before cutting a release.
 #   --yes             Accept all defaults; headless/non-interactive install
 #   --no-auto-launch  Skip automatic mosaic wizard + gateway install on first install
 #   --uninstall       Reverse the install: remove framework dir, CLI package, and npmrc line
@@ -27,6 +31,7 @@
 #   MOSAIC_PREFIX       — npm global prefix          (default: ~/.npm-global)
 #   MOSAIC_NO_COLOR     — disable colour             (set to 1)
 #   MOSAIC_REF          — git ref for framework      (default: main)
+#   MOSAIC_DEV          — equivalent to --dev         (set to 1)
 #   MOSAIC_ASSUME_YES   — equivalent to --yes        (set to 1)
 # ──────────────────────────────────────────────────────────────────────────────
 #
@@ -43,6 +48,7 @@ FLAG_CLI=true
 FLAG_NO_AUTO_LAUNCH=false
 FLAG_YES=false
 FLAG_UNINSTALL=false
+FLAG_DEV=false
 GIT_REF="${MOSAIC_REF:-main}"
 
 # MOSAIC_ASSUME_YES env var acts the same as --yes
@@ -50,12 +56,18 @@ if [[ "${MOSAIC_ASSUME_YES:-0}" == "1" ]]; then
   FLAG_YES=true
 fi
 
+# MOSAIC_DEV env var acts the same as --dev
+if [[ "${MOSAIC_DEV:-0}" == "1" ]]; then
+  FLAG_DEV=true
+fi
+
 while [[ $# -gt 0 ]]; do
   case "$1" in
     --check)          FLAG_CHECK=true; shift ;;
     --framework)      FLAG_CLI=false; shift ;;
     --cli)            FLAG_FRAMEWORK=false; shift ;;
     --ref)            GIT_REF="${2:-main}"; shift 2 ;;
+    --dev)            FLAG_DEV=true; shift ;;
     --yes|-y)         FLAG_YES=true; shift ;;
     --no-auto-launch) FLAG_NO_AUTO_LAUNCH=true; shift ;;
     --uninstall)      FLAG_UNINSTALL=true; shift ;;
@@ -72,6 +84,17 @@ CLI_PKG="${SCOPE}/mosaic"
 REPO_BASE="https://git.mosaicstack.dev/mosaicstack/stack"
 ARCHIVE_URL="${REPO_BASE}/archive/${GIT_REF}.tar.gz"
 
+# In dev (build-from-source) mode the gateway is installed globally from a
+# locally-built tarball. Tell the wizard / gateway-config stage NOT to overwrite
+# it with the registry @latest build (honored by gatewayConfigStage).
+if [[ "$FLAG_DEV" == "true" ]]; then
+  export MOSAIC_GATEWAY_SKIP_NPM_INSTALL=1
+fi
+
+# Shared monorepo checkout (populated on demand by ensure_monorepo).
+WORK_DIR=""
+EXTRACTED_DIR=""
+
 # ─── uninstall path ───────────────────────────────────────────────────────────
 # Shell-level uninstall for when the CLI is broken or not available.
 # Handles: framework directory, npm CLI package, npmrc scope line.
@@ -239,6 +262,99 @@ framework_version() {
   fi
 }
 
+# Download + extract the monorepo archive at $GIT_REF exactly once per run.
+# Sets the script-level EXTRACTED_DIR to the repo root. Reused by both the
+# framework install (Part 1) and the dev build-from-source path (Part 2).
+ensure_monorepo() {
+  if [[ -n "$EXTRACTED_DIR" ]] && [[ -d "$EXTRACTED_DIR" ]]; then
+    return 0
+  fi
+
+  require_cmd tar
+
+  WORK_DIR="$(mktemp -d "${TMPDIR:-/tmp}/mosaic-install-XXXXXX")"
+  # shellcheck disable=SC2317
+  cleanup_work() { [[ -n "$WORK_DIR" ]] && rm -rf "$WORK_DIR"; }
+  trap cleanup_work EXIT
+
+  info "Downloading source from ${GIT_REF}…"
+  if command -v curl &>/dev/null; then
+    curl -fsSL "$ARCHIVE_URL" | tar xz -C "$WORK_DIR"
+  elif command -v wget &>/dev/null; then
+    wget -qO- "$ARCHIVE_URL" | tar xz -C "$WORK_DIR"
+  else
+    fail "curl or wget required to download source."
+    exit 1
+  fi
+
+  # Gitea archives extract to <repo-name>/ inside the work dir
+  EXTRACTED_DIR="$(find "$WORK_DIR" -maxdepth 1 -mindepth 1 -type d | head -1)"
+  if [[ -z "$EXTRACTED_DIR" ]] || [[ ! -d "$EXTRACTED_DIR" ]]; then
+    fail "Could not locate extracted source in archive."
+    ls -la "$WORK_DIR" >&2
+    exit 1
+  fi
+}
+
+# Build @mosaicstack/mosaic + @mosaicstack/gateway from source and install both
+# globally from locally-packed tarballs. ZERO registry writes. Workspace deps
+# (brain/config/db/…) are pulled from the registry at the versions pinned in
+# each package.json — `pnpm pack` rewrites `workspace:*` to those versions.
+install_cli_from_source() {
+  local src="$EXTRACTED_DIR"
+  local out_dir="$WORK_DIR/dist-tarballs"
+  mkdir -p "$out_dir"
+
+  # pnpm via corepack (ships with Node >= 16.9; required by Node >= 20 preflight).
+  # Pin to the repo's packageManager version so the build matches CI. Surface
+  # corepack failures so the fresh-machine case gives an actionable error
+  # instead of a bare "command not found".
+  if ! command -v pnpm &>/dev/null; then
+    info "Activating pnpm via corepack…"
+    corepack enable 2>&1 | sed 's/^/  /' || warn "corepack enable failed — pnpm may need manual install."
+    corepack prepare pnpm@10.6.2 --activate 2>&1 | sed 's/^/  /' \
+      || warn "corepack prepare failed — pnpm may need manual install."
+  fi
+  if ! command -v pnpm &>/dev/null; then
+    fail "pnpm not available after corepack activation."
+    echo "  Install pnpm manually (https://pnpm.io/installation) and re-run with --dev."
+    exit 1
+  fi
+
+  info "Installing workspace dependencies (pnpm install)…"
+  ( cd "$src" && pnpm install ) 2>&1 | sed 's/^/  /'
+
+  info "Building CLI + gateway from source…"
+  ( cd "$src" && pnpm --filter "@mosaicstack/mosaic..." --filter "@mosaicstack/gateway..." run build ) 2>&1 | sed 's/^/  /'
+
+  info "Packing local tarballs…"
+  ( cd "$src/packages/mosaic" && pnpm pack --pack-destination "$out_dir" ) 2>&1 | sed 's/^/  /'
+  ( cd "$src/apps/gateway"    && pnpm pack --pack-destination "$out_dir" ) 2>&1 | sed 's/^/  /'
+
+  local cli_tgz gw_tgz
+  cli_tgz="$(ls -1t "$out_dir"/mosaicstack-mosaic-*.tgz 2>/dev/null | head -1)"
+  gw_tgz="$(ls -1t "$out_dir"/mosaicstack-gateway-*.tgz 2>/dev/null | head -1)"
+
+  if [[ ! -f "$cli_tgz" ]]; then
+    fail "CLI tarball was not produced by pnpm pack."
+    exit 1
+  fi
+  if [[ ! -f "$gw_tgz" ]]; then
+    fail "Gateway tarball was not produced by pnpm pack."
+    exit 1
+  fi
+
+  # Gateway first so it is present globally before the CLI's wizard runs (which
+  # skips its own gateway install via MOSAIC_GATEWAY_SKIP_NPM_INSTALL=1).
+  info "Installing gateway from source tarball (global)…"
+  npm install -g "$gw_tgz" --prefix="$PREFIX" 2>&1 | sed 's/^/  /'
+
+  info "Installing CLI from source tarball (global)…"
+  npm install -g "$cli_tgz" --prefix="$PREFIX" 2>&1 | sed 's/^/  /'
+
+  ok "Installed from source: CLI $(installed_cli_version)"
+}
+
 # ─── preflight ────────────────────────────────────────────────────────────────
 
 require_cmd node
@@ -282,25 +398,8 @@ if [[ "$FLAG_FRAMEWORK" == "true" ]]; then
       warn "Framework not installed."
     fi
   else
-    # Download repo archive and extract framework
-    require_cmd tar
-
-    WORK_DIR="$(mktemp -d "${TMPDIR:-/tmp}/mosaic-install-XXXXXX")"
-    cleanup_work() { rm -rf "$WORK_DIR"; }
-    trap cleanup_work EXIT
-
-    info "Downloading framework from ${GIT_REF}…"
-    if command -v curl &>/dev/null; then
-      curl -fsSL "$ARCHIVE_URL" | tar xz -C "$WORK_DIR"
-    elif command -v wget &>/dev/null; then
-      wget -qO- "$ARCHIVE_URL" | tar xz -C "$WORK_DIR"
-    else
-      fail "curl or wget required to download framework."
-      exit 1
-    fi
-
-    # Gitea archives extract to <repo-name>/ inside the work dir
-    EXTRACTED_DIR="$(find "$WORK_DIR" -maxdepth 1 -mindepth 1 -type d | head -1)"
+    # Download repo archive and extract framework (shared with the dev build)
+    ensure_monorepo
     FRAMEWORK_SRC="$EXTRACTED_DIR/packages/mosaic/framework"
 
     if [[ ! -d "$FRAMEWORK_SRC" ]]; then
@@ -356,7 +455,11 @@ if [[ "$FLAG_CLI" == "true" ]]; then
   fi
 
   CURRENT="$(installed_cli_version)"
-  LATEST="$(latest_cli_version)"
+  if [[ "$FLAG_DEV" == "true" ]]; then
+    LATEST=""
+  else
+    LATEST="$(latest_cli_version)"
+  fi
 
   if [[ -n "$CURRENT" ]]; then
     dim "  Installed: ${CLI_PKG}@${CURRENT}"
@@ -364,7 +467,9 @@ if [[ "$FLAG_CLI" == "true" ]]; then
     dim "  Installed: (none)"
   fi
 
-  if [[ -n "$LATEST" ]]; then
+  if [[ "$FLAG_DEV" == "true" ]]; then
+    dim "  Source:    ${REPO_BASE} (ref: ${GIT_REF}, build-from-source)"
+  elif [[ -n "$LATEST" ]]; then
     dim "  Latest:    ${CLI_PKG}@${LATEST}"
   else
     dim "  Latest:    (registry unreachable)"
@@ -372,7 +477,9 @@ if [[ "$FLAG_CLI" == "true" ]]; then
   echo ""
 
   if [[ "$FLAG_CHECK" == "true" ]]; then
-    if [[ -z "$LATEST" ]]; then
+    if [[ "$FLAG_DEV" == "true" ]]; then
+      info "Dev mode: installed version is ${CURRENT:-(none)} (no registry comparison)."
+    elif [[ -z "$LATEST" ]]; then
       warn "Could not reach registry."
     elif [[ -z "$CURRENT" ]]; then
       warn "Not installed."
@@ -383,6 +490,16 @@ if [[ "$FLAG_CLI" == "true" ]]; then
     else
       ok "Up to date (or ahead of registry)."
     fi
+  elif [[ "$FLAG_DEV" == "true" ]]; then
+    info "Dev mode — building CLI + gateway from source at ref ${GIT_REF}…"
+    ensure_monorepo
+    install_cli_from_source
+
+    # PATH check for npm prefix
+    if [[ ":$PATH:" != *":$PREFIX/bin:"* ]]; then
+      warn "$PREFIX/bin is not on your PATH"
+      dim "  Add to your shell rc:  export PATH=\"$PREFIX/bin:\$PATH\""
+    fi
   else
     if [[ -z "$LATEST" ]]; then
       warn "Could not reach registry at $REGISTRY — skipping npm CLI."