docs(design): durable resume state for session restart

Full mission status: open PRs (#543, #570), locked decisions, P0-P6 phase
status, drift re-grounding, and the step-by-step RESUME PROCEDURE.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

.env.example

@@ -1,24 +1,154 @@
-# Non-secret runtime settings for the mosaic-poc-agent container.
-# Copy to .env if you want to override the defaults in compose.yaml.
-#
-# NEVER put credentials in this file. Authentication is supplied at
-# runtime only, via one of the two documented paths:
-#   1. read-only mounted pi auth file (default: ~/.pi/agent/auth.json,
-#      override the host path with PI_AUTH_FILE)
-#   2. provider API key environment variable (ZAI_API_KEY or
-#      ANTHROPIC_API_KEY), passed through by compose.yaml when set
+# ─────────────────────────────────────────────────────────────────────────────
+# Mosaic — Environment Variables Reference
+# Copy this file to .env and fill in the values for your deployment.
+# Lines beginning with # are comments; optional vars are commented out.
+# ─────────────────────────────────────────────────────────────────────────────
 
-# Model provider (built-in pi provider name)
-PI_PROVIDER=zai
 
-# Model ID within the provider
-PI_MODEL=glm-5.3-flash
+# ─── Database (PostgreSQL 17 + pgvector) ─────────────────────────────────────
+# Full connection string used by the gateway, ORM, and migration runner.
+# Port 5433 avoids conflict with a host-side PostgreSQL instance.
+DATABASE_URL=postgresql://mosaic:mosaic@localhost:5433/mosaic
 
-# Optional: alternative host path of the pi credential file mounted
-# read-only at /home/node/.pi/agent/auth.json in the container
-#PI_AUTH_FILE=/home/jwoltje/.pi/agent/auth.json
+# Docker Compose host-port override for the PostgreSQL container (default: 5433)
+# PG_HOST_PORT=5433
 
-# Optional: documented env-var auth alternative (secret! set in your
-# shell or a gitignored .env, never commit)
-#ZAI_API_KEY=
-#ANTHROPIC_API_KEY=
+
+# ─── Queue (Valkey 8 / Redis-compatible) ─────────────────────────────────────
+# Port 6380 avoids conflict with a host-side Redis/Valkey instance.
+VALKEY_URL=redis://localhost:6380
+
+# Docker Compose host-port override for the Valkey container (default: 6380)
+# VALKEY_HOST_PORT=6380
+
+
+# ─── Gateway ─────────────────────────────────────────────────────────────────
+# TCP port the NestJS/Fastify gateway listens on (default: 14242)
+GATEWAY_PORT=14242
+
+# Comma-separated list of allowed CORS origins.
+# Must include the web app origin in production.
+GATEWAY_CORS_ORIGIN=http://localhost:3000
+
+
+# ─── Auth (BetterAuth) ───────────────────────────────────────────────────────
+# REQUIRED — random secret used to sign sessions and tokens.
+# Generate with: openssl rand -base64 32
+BETTER_AUTH_SECRET=change-me-to-a-random-32-char-string
+
+# Public base URL of the gateway (used by BetterAuth for callback URLs)
+BETTER_AUTH_URL=http://localhost:14242
+
+
+# ─── Web App (Next.js) ───────────────────────────────────────────────────────
+# Public gateway URL — accessible from the browser, not just the server.
+NEXT_PUBLIC_GATEWAY_URL=http://localhost:14242
+
+
+# ─── OpenTelemetry ───────────────────────────────────────────────────────────
+# OTLP HTTP endpoint (otel-collector or any OpenTelemetry-compatible backend)
+OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
+
+# Service name shown in traces
+OTEL_SERVICE_NAME=mosaic-gateway
+
+
+# ─── AI Providers ────────────────────────────────────────────────────────────
+
+# Ollama (local models — set OLLAMA_BASE_URL to enable)
+# OLLAMA_BASE_URL=http://localhost:11434
+# OLLAMA_HOST is a legacy alias for OLLAMA_BASE_URL
+# OLLAMA_HOST=http://localhost:11434
+# Comma-separated list of Ollama model IDs to register (default: llama3.2,codellama,mistral)
+# OLLAMA_MODELS=llama3.2,codellama,mistral
+
+# Anthropic (claude-sonnet-4-6, claude-opus-4-6, claude-haiku-4-5)
+# ANTHROPIC_API_KEY=sk-ant-...
+
+# OpenAI (gpt-4o, gpt-4o-mini, o3-mini)
+# OPENAI_API_KEY=sk-...
+
+# Z.ai / GLM (glm-4.5, glm-4.5-air, glm-4.5-flash)
+# ZAI_API_KEY=...
+
+# Custom providers — JSON array of provider configs
+# Format: [{"id":"<id>","baseUrl":"<url>","apiKey":"<key>","models":[{"id":"<model-id>","name":"<label>"}]}]
+# MOSAIC_CUSTOM_PROVIDERS=
+
+
+# ─── Embedding Service ───────────────────────────────────────────────────────
+# OpenAI-compatible embeddings endpoint (default: OpenAI)
+# EMBEDDING_API_URL=https://api.openai.com/v1
+# EMBEDDING_MODEL=text-embedding-3-small
+
+
+# ─── Log Summarization Service ───────────────────────────────────────────────
+# OpenAI-compatible chat completions endpoint for log summarization (default: OpenAI)
+# SUMMARIZATION_API_URL=https://api.openai.com/v1
+# SUMMARIZATION_MODEL=gpt-4o-mini
+
+# Cron schedule for summarization job (default: every 6 hours)
+# SUMMARIZATION_CRON=0 */6 * * *
+
+# Cron schedule for log tier management (default: daily at 03:00)
+# TIER_MANAGEMENT_CRON=0 3 * * *
+
+
+# ─── Agent ───────────────────────────────────────────────────────────────────
+# Filesystem sandbox root for agent file tools (default: process.cwd())
+# AGENT_FILE_SANDBOX_DIR=/var/lib/mosaic/sandbox
+
+# Comma-separated list of tool names available to non-admin users.
+# Leave unset to allow all tools for all authenticated users.
+# AGENT_USER_TOOLS=read_file,list_directory,search_files
+
+# System prompt injected into every agent session (optional)
+# AGENT_SYSTEM_PROMPT=You are a helpful assistant.
+
+
+# ─── MCP Servers ─────────────────────────────────────────────────────────────
+# JSON array of MCP server configs — set to enable MCP tool integration.
+# Each entry: {"name":"<id>","url":"<http-or-sse-url>"}
+# MCP_SERVERS=[{"name":"my-mcp","url":"http://localhost:3100/sse"}]
+
+
+# ─── Coordinator ─────────────────────────────────────────────────────────────
+# Root directory used to scope coordinator (worktree/repo) operations.
+# Defaults to the monorepo root auto-detected from process.cwd().
+# MOSAIC_WORKSPACE_ROOT=/home/user/projects/mosaic
+
+
+# ─── Discord Plugin (optional — set DISCORD_BOT_TOKEN to enable) ─────────────
+# DISCORD_BOT_TOKEN=
+# DISCORD_GUILD_ID=
+# DISCORD_GATEWAY_URL=http://localhost:14242
+
+
+# ─── Telegram Plugin (optional — set TELEGRAM_BOT_TOKEN to enable) ───────────
+# TELEGRAM_BOT_TOKEN=
+# TELEGRAM_GATEWAY_URL=http://localhost:14242
+
+
+# ─── SSO Providers (add credentials to enable) ───────────────────────────────
+
+# --- Authentik (optional — set AUTHENTIK_CLIENT_ID to enable) ---
+# AUTHENTIK_ISSUER=https://auth.example.com/application/o/mosaic/
+# AUTHENTIK_CLIENT_ID=
+# AUTHENTIK_CLIENT_SECRET=
+
+# --- WorkOS (optional — set WORKOS_CLIENT_ID to enable) ---
+# WORKOS_ISSUER=https://your-company.authkit.app
+# WORKOS_CLIENT_ID=client_...
+# WORKOS_CLIENT_SECRET=sk_live_...
+
+# --- Keycloak (optional — set KEYCLOAK_CLIENT_ID to enable) ---
+# KEYCLOAK_ISSUER=https://auth.example.com/realms/master
+# Legacy alternative if you prefer to compose the issuer from separate vars:
+# KEYCLOAK_URL=https://auth.example.com
+# KEYCLOAK_REALM=master
+# KEYCLOAK_CLIENT_ID=mosaic
+# KEYCLOAK_CLIENT_SECRET=
+
+# Feature flags — set to true alongside provider credentials to show SSO buttons in the UI
+# NEXT_PUBLIC_WORKOS_ENABLED=true
+# NEXT_PUBLIC_KEYCLOAK_ENABLED=true

.gitignore

@@ -1,8 +1,14 @@
-# build/deps
-node_modules/
-
-# runtime credentials — never commit, never copy into the image
+logs/
+node_modules
+dist
+.turbo
+.next
+coverage
 .env
-secrets/
+.env.local
+*.tsbuildinfo
+.pnpm-store
+docs/reports/
 
-# generated runtime state lives in /home/jwoltje/.mosaic-dev (outside this project)
+# Step-CA dev password — real file is gitignored; commit only the .example
+infra/step-ca/dev-password

.husky/pre-push

@@ -0,0 +1 @@
+pnpm typecheck && pnpm lint && pnpm format:check

.mosaic/project.json

@@ -1,12 +0,0 @@
-{
-  "projectVersion": 1,
-  "id": "stack",
-  "vars": {
-    "tracker.project": 19,
-    "git.workingBranch": "refactor",
-    "git.protectedBranches": [
-      "main",
-      "next"
-    ]
-  }
-}

.npmrc

@@ -0,0 +1 @@
+@mosaicstack:registry=https://git.mosaicstack.dev/api/packages/mosaicstack/npm/

.pi/.gitignore

@@ -1,6 +0,0 @@
-extensions/
-extensions.installed.sha256
-.extensions-*
-state/
-evidence/
-native-test-*.log

.pi/README.md

@@ -1,34 +0,0 @@
-# Native goal development copy
-
-From this repository, start a fresh native Pi session:
-
-```sh
-bash scripts/goal-dev.sh
-```
-
-Canonical source lives under `extensions/`. The launcher first runs `scripts/sync-dev-extensions.sh`, which installs verified ordinary-file copies under `.pi/extensions/`, then loads only the generated goal extension. Global extensions remain unloaded. The launcher keeps your usual native Pi provider authentication; it copies no credentials. Goal state and new conversation files live under `.pi/state/`, which is ignored by Git. Each process gets a fresh incarnation; `/reload` and `/new` in the same process retain its goal. Restarting Pi does not adopt an earlier process's active goal.
-
-Plain `pi` also discovers `.pi/extensions/goal/index.ts` after project trust, but may load global extensions too. Use the launcher to avoid duplicate `/goal` registrations. This is a local development test, not a sandbox or the managed Mosaic runtime. Docker and `~/.mosaic` are unchanged.
-
-## Try it
-
-1. Set `/goal <a long goal with acceptance criteria>`. This starts work immediately.
-2. Look below the editor for `Goal: Active`. The old above-editor goal widget is gone.
-3. Run bare `/goal`, then press `Alt+G`. Both show the entire stored goal and its status. Tab remains autocomplete.
-4. Use `/goal stop` and `/goal resume`. Expect Paused and Active, or Waiting if an untimed wait remains recorded.
-5. A blocked `goal_report` displays Blocked. A satisfied report displays Complete and retains the full goal for recall without continuing work.
-6. `/goal clear` removes the retained goal. Try `NO_COLOR=1 bash scripts/goal-dev.sh` to check text-only labels.
-
-Use terminal scrollback for recall longer than the screen. At narrow widths Pi may truncate its footer status row; bare `/goal` and Alt+G remain available.
-
-## Checks
-
-```sh
-node --test extensions/goal/test/*.test.ts
-bash scripts/test-extension-package.sh
-python3 scripts/test-goal-native.py
-```
-
-Contract tests use ordinary read-only fixture copies in `test/fixtures/skills-local/`, not live brain files. The executive-update fixture SHA-256 matches the parser's pinned contract, `bbea48a46b1f8da7bc759f86856fb52830b7dde456b826317163c6dc6ccab319`.
-
-`SOURCE-SNAPSHOT.json` records the original external-source baseline, not the edited candidate. No symlinks are used. Never edit `.pi/extensions/`; the sync script refuses to overwrite installation drift. Make changes under `extensions/`, run the checks, and relaunch. To disable the test, stop its Pi process and remove `.pi/extensions/`. Keep `.pi/state/` only if you need local test state.

.pi/SOURCE-SNAPSHOT.json

@@ -1,10 +0,0 @@
-{
-  "snapshotVersion": 1,
-  "copiedAt": "2026-09-06T04:58:22Z",
-  "source": "~/.mosaic/fleet/extensions",
-  "goalTreeSha256": "8853f2b72dde3e87c4573648b9a931c1c75da87ccde995c3224e6d2e707a75f0",
-  "mosaicCoreLibTreeSha256": "d1194dce31209e5773c6cc5ce571cbca3c39b29d943a79dea06665e05d29f319",
-  "symlinks": false,
-  "autoDiscoveredExtensions": ["goal"],
-  "purpose": "Issue #54 native Pi NG development copy; never loaded by Docker"
-}

.pi/goal-dev.sh

@@ -1,5 +0,0 @@
-#!/usr/bin/env bash
-# Compatibility entrypoint for the accepted native test command.
-set -euo pipefail
-cd "$(dirname "${BASH_SOURCE[0]}")/.."
-exec scripts/goal-dev.sh "$@"

.prettierignore

@@ -0,0 +1,7 @@
+pnpm-lock.yaml
+**/next-env.d.ts
+**/dist
+**/node_modules
+**/drizzle
+**/.next
+.claude/

.woodpecker/ci.yml

@@ -0,0 +1,84 @@
+variables:
+  - &node_image 'node:22-alpine'
+  - &enable_pnpm 'corepack enable'
+
+when:
+  - event: [push, pull_request, manual]
+
+# Turbo remote cache (turbo.mosaicstack.dev) is configured via Woodpecker
+# repository-level environment variables (TURBO_API, TURBO_TEAM, TURBO_TOKEN).
+# This avoids from_secret which is blocked on pull_request events.
+# If the env vars aren't set, turbo falls back to local cache only.
+
+steps:
+  install:
+    image: *node_image
+    commands:
+      - corepack enable
+      - apk add --no-cache python3 make g++
+      - pnpm install --frozen-lockfile
+
+  typecheck:
+    image: *node_image
+    commands:
+      - *enable_pnpm
+      - pnpm typecheck
+    depends_on:
+      - install
+
+  # lint, format, and test are independent — run in parallel after typecheck
+  lint:
+    image: *node_image
+    commands:
+      - *enable_pnpm
+      - pnpm lint
+    depends_on:
+      - typecheck
+
+  format:
+    image: *node_image
+    commands:
+      - *enable_pnpm
+      - pnpm format:check
+    depends_on:
+      - typecheck
+
+  test:
+    image: *node_image
+    environment:
+      # Avoid the namespace-level Woodpecker DB service named "postgres".
+      # The Kubernetes backend exposes service containers by step name.
+      DATABASE_URL: postgresql://mosaic:mosaic@ci-postgres:5432/mosaic
+    commands:
+      - *enable_pnpm
+      # Install postgresql-client for pg_isready
+      - apk add --no-cache postgresql-client
+      # Wait up to 60s for CI postgres to be ready; fail fast if it never comes up.
+      - |
+        ready=0
+        for i in $(seq 1 60); do
+          if pg_isready -h ci-postgres -p 5432 -U mosaic; then
+            ready=1
+            break
+          fi
+          echo "Waiting for ci-postgres ($i/60)..."
+          sleep 1
+        done
+        if [ "$ready" -ne 1 ]; then
+          echo "ci-postgres did not become ready" >&2
+          exit 1
+        fi
+      # Run migrations (DATABASE_URL is set in environment above)
+      - pnpm --filter @mosaicstack/db run db:migrate
+      # Run all tests
+      - pnpm test
+    depends_on:
+      - typecheck
+
+services:
+  ci-postgres:
+    image: pgvector/pgvector:pg17
+    environment:
+      POSTGRES_USER: mosaic
+      POSTGRES_PASSWORD: mosaic
+      POSTGRES_DB: mosaic

.woodpecker/publish.yml

@@ -0,0 +1,165 @@
+# Build, publish npm packages, and push Docker images
+# Runs only on main branch push/tag
+
+variables:
+  - &node_image 'node:22-alpine'
+  - &enable_pnpm 'corepack enable'
+
+when:
+  - branch: [main]
+    event: [push, manual, tag]
+
+steps:
+  install:
+    image: *node_image
+    commands:
+      - corepack enable
+      - pnpm install --frozen-lockfile
+
+  build:
+    image: *node_image
+    commands:
+      - *enable_pnpm
+      - pnpm build
+    depends_on:
+      - install
+
+  publish-npm:
+    image: *node_image
+    environment:
+      NPM_TOKEN:
+        from_secret: gitea_token
+    commands:
+      - *enable_pnpm
+      # Configure auth for Gitea npm registry
+      - |
+        echo "//git.mosaicstack.dev/api/packages/mosaicstack/npm/:_authToken=$NPM_TOKEN" > ~/.npmrc
+        echo "@mosaicstack:registry=https://git.mosaicstack.dev/api/packages/mosaicstack/npm/" >> ~/.npmrc
+      # Publish non-private packages to Gitea.
+      #
+      # The only publish failure we tolerate is "version already exists" —
+      # that legitimately happens when only some packages were bumped in
+      # the merge. Any other failure (registry 404, auth error, network
+      # error) MUST fail the pipeline loudly: the previous
+      # `|| echo "... continuing"` fallback silently hid a 404 from the
+      # Gitea org rename and caused every @mosaicstack/* publish to fall
+      # on the floor while CI still reported green.
+      - |
+        # Portable sh (Alpine ash) — avoid bashisms like PIPESTATUS.
+        set +e
+        pnpm --filter "@mosaicstack/*" --filter "!@mosaicstack/web" publish --no-git-checks --access public >/tmp/publish.log 2>&1
+        EXIT=$?
+        set -e
+        cat /tmp/publish.log
+        if [ "$EXIT" -eq 0 ]; then
+          echo "[publish] all packages published successfully"
+          exit 0
+        fi
+        # Hard registry / auth / network errors → fatal. Match npm's own
+        # error lines specifically to avoid false positives on arbitrary
+        # log text that happens to contain "E404" etc.
+        if grep -qE "npm (error|ERR!) code (E404|E401|ENEEDAUTH|ECONNREFUSED|ETIMEDOUT|ENOTFOUND)" /tmp/publish.log; then
+          echo "[publish] FATAL: registry/auth/network error detected — failing pipeline" >&2
+          exit 1
+        fi
+        # Only tolerate the explicit "version already published" case.
+        # npm returns this as E403 with body "You cannot publish over..."
+        # or EPUBLISHCONFLICT depending on version.
+        if grep -qE "EPUBLISHCONFLICT|You cannot publish over|previously published" /tmp/publish.log; then
+          echo "[publish] some packages already at this version — continuing (non-fatal)"
+          exit 0
+        fi
+        echo "[publish] FATAL: publish failed with unrecognized error — failing pipeline" >&2
+        exit 1
+    depends_on:
+      - build
+
+  # TODO: Uncomment when ready to publish to npmjs.org
+  # publish-npmjs:
+  #   image: *node_image
+  #   environment:
+  #     NPM_TOKEN:
+  #       from_secret: npmjs_token
+  #   commands:
+  #     - *enable_pnpm
+  #     - apk add --no-cache jq bash
+  #     - bash scripts/publish-npmjs.sh
+  #   depends_on:
+  #     - build
+  #   when:
+  #     - event: [tag]
+
+  build-gateway:
+    image: gcr.io/kaniko-project/executor:debug
+    environment:
+      REGISTRY_USER:
+        from_secret: gitea_username
+      REGISTRY_PASS:
+        from_secret: gitea_password
+      CI_COMMIT_BRANCH: ${CI_COMMIT_BRANCH}
+      CI_COMMIT_TAG: ${CI_COMMIT_TAG}
+      CI_COMMIT_SHA: ${CI_COMMIT_SHA}
+    commands:
+      - mkdir -p /kaniko/.docker
+      - echo "{\"auths\":{\"git.mosaicstack.dev\":{\"username\":\"$REGISTRY_USER\",\"password\":\"$REGISTRY_PASS\"}}}" > /kaniko/.docker/config.json
+      - |
+        DESTINATIONS="--destination git.mosaicstack.dev/mosaicstack/stack/gateway:sha-${CI_COMMIT_SHA:0:7}"
+        if [ "$CI_COMMIT_BRANCH" = "main" ]; then
+          DESTINATIONS="$DESTINATIONS --destination git.mosaicstack.dev/mosaicstack/stack/gateway:latest"
+        fi
+        if [ -n "$CI_COMMIT_TAG" ]; then
+          DESTINATIONS="$DESTINATIONS --destination git.mosaicstack.dev/mosaicstack/stack/gateway:$CI_COMMIT_TAG"
+        fi
+        /kaniko/executor --context . --dockerfile docker/gateway.Dockerfile $DESTINATIONS
+    depends_on:
+      - build
+
+  build-appservice:
+    image: gcr.io/kaniko-project/executor:debug
+    environment:
+      REGISTRY_USER:
+        from_secret: gitea_username
+      REGISTRY_PASS:
+        from_secret: gitea_password
+      CI_COMMIT_BRANCH: ${CI_COMMIT_BRANCH}
+      CI_COMMIT_TAG: ${CI_COMMIT_TAG}
+      CI_COMMIT_SHA: ${CI_COMMIT_SHA}
+    commands:
+      - mkdir -p /kaniko/.docker
+      - echo "{\"auths\":{\"git.mosaicstack.dev\":{\"username\":\"$REGISTRY_USER\",\"password\":\"$REGISTRY_PASS\"}}}" > /kaniko/.docker/config.json
+      - |
+        DESTINATIONS="--destination git.mosaicstack.dev/mosaicstack/stack/appservice:sha-${CI_COMMIT_SHA:0:7}"
+        if [ "$CI_COMMIT_BRANCH" = "main" ]; then
+          DESTINATIONS="$DESTINATIONS --destination git.mosaicstack.dev/mosaicstack/stack/appservice:latest"
+        fi
+        if [ -n "$CI_COMMIT_TAG" ]; then
+          DESTINATIONS="$DESTINATIONS --destination git.mosaicstack.dev/mosaicstack/stack/appservice:$CI_COMMIT_TAG"
+        fi
+        /kaniko/executor --context . --dockerfile docker/appservice.Dockerfile $DESTINATIONS
+    depends_on:
+      - build
+
+  build-web:
+    image: gcr.io/kaniko-project/executor:debug
+    environment:
+      REGISTRY_USER:
+        from_secret: gitea_username
+      REGISTRY_PASS:
+        from_secret: gitea_password
+      CI_COMMIT_BRANCH: ${CI_COMMIT_BRANCH}
+      CI_COMMIT_TAG: ${CI_COMMIT_TAG}
+      CI_COMMIT_SHA: ${CI_COMMIT_SHA}
+    commands:
+      - mkdir -p /kaniko/.docker
+      - echo "{\"auths\":{\"git.mosaicstack.dev\":{\"username\":\"$REGISTRY_USER\",\"password\":\"$REGISTRY_PASS\"}}}" > /kaniko/.docker/config.json
+      - |
+        DESTINATIONS="--destination git.mosaicstack.dev/mosaicstack/stack/web:sha-${CI_COMMIT_SHA:0:7}"
+        if [ "$CI_COMMIT_BRANCH" = "main" ]; then
+          DESTINATIONS="$DESTINATIONS --destination git.mosaicstack.dev/mosaicstack/stack/web:latest"
+        fi
+        if [ -n "$CI_COMMIT_TAG" ]; then
+          DESTINATIONS="$DESTINATIONS --destination git.mosaicstack.dev/mosaicstack/stack/web:$CI_COMMIT_TAG"
+        fi
+        /kaniko/executor --context . --dockerfile docker/web.Dockerfile $DESTINATIONS
+    depends_on:
+      - build

AGENTS.md

@@ -1,222 +1,80 @@
-# AGENTS.md — Mosaic Stack rebuild (`mosaicstack/stack`, branch `refactor`)
+# Agent Guidelines — Mosaic Stack
 
-Operational context for any agent session working in this repository.
-Read top to bottom; it is deliberately short — depth lives in the files it
-points to, not here.
+## Required Load Order
 
-## What this repository is
+1. `~/.config/mosaic/SOUL.md`
+2. `~/.config/mosaic/STANDARDS.md`
+3. `~/.config/mosaic/AGENTS.md`
+4. `~/.config/mosaic/guides/E2E-DELIVERY.md`
+5. `AGENTS.md` (this file)
+6. Runtime-specific guide: `~/.config/mosaic/runtime/<runtime>/RUNTIME.md`
 
-Canonical checkout: `/mnt/storage/src/mosaic-stack`, origin `mosaicstack/stack`,
-working branch `refactor` (Jason-authorized conversion, issue #1495).
-The new foundation is at the root. `v1/` is archived legacy source, not the current
-implementation; its instructions and tools do not govern the new foundation.
-`~/src/mosaic-stack-dev-test` is a compatibility symlink to this checkout, not a
-second working tree. Both original Git histories are retained. Conversion receipt:
-`docs/plans/2026-09-07_repository-consolidation-completed.md`.
+## Project Context
 
-A rebuild of Mosaic Stack: a file-based, fail-closed
-orchestration foundation that dispatches sandboxed headless pi workers to do
-real work, with immutable run records as evidence. Thirteen-plus tagged
-milestones (`git tag -l`) from `poc-container-hello-v0` to today; suites
-green at every step. Not production software — a proven foundation.
+Mosaic Stack is a self-hosted, multi-user AI agent platform. TypeScript monorepo with NestJS gateway, Next.js web dashboard, Pi SDK agent runtime, and plugin architecture for Discord/Telegram.
 
-## Non-negotiable invariants (the canon)
+## Package Map
 
-1. **Root is bootstrap-only.** First-class system configuration lives at the
-   repository root; everything else gets a dedicated directory (`roles/`,
-   `contracts/`, `missions/`, `tasks/`, `docs/`). Do not add new files to root.
-2. **Configuration**: each stack instance has exactly one system config. The
-   default instance's is `~/.config/mosaic-dev/config.json`. Instance NAME's
-   is `~/.config/mosaic-NAME/config.json`, and its data root carries a marker
-   that names it. `~` is the home of the OS user that runs the instance. A
-   process reads only the config `MOSAIC_CONFIG` names, or the default's when
-   that is unset. Configs are created only by
-   `scripts/bootstrap.sh [--instance NAME]`, never overwritten, fail-closed on
-   any problem. Repo-scoped role authority lives in `roles/*.json`
-   (versioned, reviewed commits only).
-3. **Secrets** never enter the repository or container images; auth is
-   runtime-only (read-only mount or environment variable).
-4. **Contracts** (`contracts/`) are immutable and image-baked. Missions and
-   tasks are declarative JSON with strict schemas.
-5. **Run records** under `<dataRoot>/runs/` are write-once evidence — never
-   rewritten, only pruned via `prune` with a receipt.
-6. **Fail closed**: missing or invalid config/policy refuses the operation.
-   Never improvise around a refusal; diagnose it.
-7. **Policy**: missions govern tasks (least-privilege intersection — a task
-   narrows, never widens). Role authority is declared in `roles/` and changes
-   only via reviewed commits.
-8. **Git**: commit only after applicable suites are green. Work on the
-   owner-authorized `refactor` branch; never force-push. Push remains an explicit
-   act. Do not merge into `next` or `main` without separate authorization.
-   `scripts/conductor-apply.sh` commits locally; it does not authorize a push.
-9. **Append-only logs**: BUILD-LOG.md (phases), `activation-log.jsonl`,
-   `.pruned.log`, docs/SESSIONS.md. Corrections are new entries, never edits.
+| Package            | Purpose                         | Key Dependencies                 |
+| ------------------ | ------------------------------- | -------------------------------- |
+| `apps/gateway`     | NestJS API + WebSocket hub      | Fastify, Socket.IO, Pi SDK, OTEL |
+| `apps/web`         | Next.js dashboard               | React 19, Tailwind               |
+| `packages/types`   | Shared TypeScript contracts     | class-validator                  |
+| `packages/db`      | Drizzle ORM schema + migrations | drizzle-orm, postgres            |
+| `packages/auth`    | BetterAuth configuration        | better-auth, @mosaicstack/db     |
+| `packages/brain`   | Data layer (PG-backed)          | @mosaicstack/db                  |
+| `packages/queue`   | Valkey task queue + MCP         | ioredis                          |
+| `packages/coord`   | Mission coordination            | @mosaicstack/queue               |
+| `packages/mosaic`  | Unified `mosaic` CLI + TUI      | Ink, Pi SDK, commander           |
+| `plugins/discord`  | Discord channel plugin          | discord.js                       |
+| `plugins/telegram` | Telegram channel plugin         | Telegraf                         |
 
-## Autonomous operation within an agreed plan
+## Architecture Rules
 
-Autonomy starts after alignment, not before it. For a new substantial assignment,
-recover the applicable mission, goal, task, `CURRENT.md` state, and prior owner
-decisions, then work with the user to establish a plan of action: the intended
-outcome, acceptance evidence, boundaries, and any gated actions. Recommend a
-concrete plan instead of presenting an open-ended menu. A direct request or
-existing approved plan that already settles those points is sufficient alignment;
-do not ask for ceremonial reconfirmation.
+1. Gateway is the single API surface — all clients connect through it
+2. Pi SDK is ESM-only — gateway and CLI must use ESM
+3. Socket.IO typed events defined in `@mosaicstack/types` enforce compile-time contracts
+4. OTEL auto-instrumentation loads before NestJS bootstrap
+5. BetterAuth manages auth tables; schema defined in `@mosaicstack/db`
+6. Docker Compose provides PG (5433), Valkey (6380), OTEL Collector (4317/4318), Jaeger (16686)
+7. Explicit `@Inject()` decorators required in NestJS (tsx/esbuild doesn't emit decorator metadata)
 
-Once the plan is established, carry it to verified completion without prompting
-for routine decisions or permission to take the next in-scope step. Authorization
-persists for the life of that assignment unless the user changes or revokes it.
-Treat mid-session user input as steering: incorporate it, update the plan or
-tracking record when needed, and continue.
+## Development Workflow
 
-### Decide and continue
+```bash
+docker compose up -d          # Infrastructure
+pnpm install                  # Dependencies
+pnpm typecheck && pnpm lint && pnpm format:check  # Quality gates
+```
 
-- Resolve naming, implementation approach, layout, and similar non-breaking
-  choices from, in order: repository invariants and role policy, the approved
-  plan and acceptance criteria, established repository conventions, then the
-  smallest reversible option. Record a consequential choice and its tradeoff.
-- Perform the in-scope investigation, edits, tests, documentation, and tracking
-  needed for end-to-end acceptance. Do not ask whether to add obviously required
-  tests or documentation.
-- Diagnose failures and retry or remediate within the agreed scope. Fix a defect
-  when it blocks acceptance or is local to files already being changed; otherwise
-  record a bounded follow-up without expanding the assignment.
-- Resolve minor ambiguity in favor of the mission, goal, north star, and prior
-  owner decisions. State the assumption in the completion report.
-- Never stop merely to ask whether to proceed, which routine option to use, or
-  whether to execute the next step already contained in the plan.
+## Repo-Specific Notes
 
-### Re-align or stop only at a real boundary
+- DTOs in `*.dto.ts` files at module boundaries
+- ESM everywhere (`"type": "module"`, `.js` extensions in imports)
+- NodeNext module resolution in all tsconfigs
+- Scratchpads are mandatory for non-trivial tasks
 
-Finish all independent work first, then ask one focused question only when:
+## docs/TASKS.md — Schema (CANONICAL)
 
-1. Two plausible readings materially change the outcome and the choice is costly
-   to reverse.
-2. The next action would exceed the agreed scope or authority, introduce an
-   unapproved breaking public/API/schema/data/policy change, or alter a security
-   boundary.
-3. Credentials or access are missing and no in-scope path remains.
-4. The action is destructive, irreversible, production-affecting, incurs spend,
-   or communicates externally on the user's behalf without explicit authority.
-5. Objectives or owner decisions genuinely conflict and repository evidence
-   cannot resolve them.
-6. A fail-closed policy refusal or another agent's overlapping ownership prevents
-   safe progress. Diagnose and report it; never route around it.
+The `agent` column specifies the required model for each task. **This is set at task creation by the orchestrator and must not be changed by workers.**
 
-Repository gates still apply. In particular, a successful implementation or a
-broad request to “finish” does not by itself authorize push, merge, deployment,
-release, production changes, policy/role expansion, or access to secrets. Perform
-such an action only when the established plan explicitly includes it. If blocked,
-report the exact boundary, what is complete, the recommended resolution, and the
-specific action that will resume; do not use “waiting for confirmation” as a
-substitute for a real blocker.
+| Value     | When to use                                                 | Budget                     |
+| --------- | ----------------------------------------------------------- | -------------------------- |
+| `codex`   | All coding tasks (default for implementation)               | OpenAI credits — preferred |
+| `glm-5.1` | Cost-sensitive coding where Codex is unavailable            | Z.ai credits               |
+| `haiku`   | Review gates, verify tasks, status checks, docs-only        | Cheapest Claude tier       |
+| `sonnet`  | Complex planning, multi-file reasoning, architecture review | Claude quota               |
+| `opus`    | Major cross-cutting architecture decisions ONLY             | Most expensive — minimize  |
+| `—`       | No preference / auto-select cheapest capable                | Pipeline decides           |
 
-## Session protocol (mandatory)
+Pipeline crons read this column and spawn accordingly. Workers never modify `docs/TASKS.md` — only the orchestrator writes it.
 
-- **Register** your session in `docs/SESSIONS.md` — one append-only line
-  (date, actor, scope, outcome). Never rewrite or remove entries.
-- **Cadence**: run `scripts/mosaic queue next <your seat>` first. It names
-  the row to resume, review or start, or says there is nothing. The goal order
-  in `docs/plans/2026-09-27_goals-review.md` sets priority, not CURRENT.md.
-  Open only the brief that row links to. Execute it through every authorized
-  stage (implement → test → verify against acceptance criteria; commit, push,
-  or close only when the established plan authorizes each) → move the row with
-  `scripts/mosaic queue move` (never by editing QUEUE.md) → register in
-  SESSIONS.md.
-- "next" means one action. A batch mandate ("run the queue") repeats the
-  loop until green or truly blocked under the boundary rules above.
-- Substantial work gets a Gitea issue and a BUILD-LOG phase entry
-  (before/after, with corrections recorded honestly).
+**Full schema:**
 
-## Internal development bootstrap
+```
+| id | status | description | issue | agent | repo | branch | depends_on | estimate | notes |
+```
 
-Jason's current direction is repository-native development in
-`/mnt/storage/src/mosaic-stack`. Sage leads the project (Jason's ruling,
-2026-09-26) and coordinates coding, review and research through Darkwing, Dewey,
-Filbert, Rocko, Researcher and any further seats Jason launches under `agents/`.
-Darkwing is a collaborating agent seat, not the coordinator. Development sessions
-run in T3 for now. Work moves to the new stack; the old `~/.mosaic` fleet is being
-retired, and a fleet seat acting outside Jason's instructions is the failure this
-transition exists to prevent.
-Do not assign new development work to fleet seats during this bootstrap phase.
-Do not modify `~/.mosaic` launchers, provisioning or other state, or stop/migrate
-live fleet processes as part of this work. Preserve existing work and histories.
-Use the repository bootstrap/configuration and launch entry points; missing
-configuration still fails closed. This changes development coordination, not
-managed worker role policy or deployment authority. The lead role adds no push,
-merge or deployment authority; those still need Jason's say-so. See
-`agents/README.md` for the internal roster.
-
-For control-board attention, start a completed reply with `Input needed: ` and
-one specific nonempty request only when Jason must provide a decision or input.
-Put that line at column zero, before other text. Do not use it for routine
-completion or a wait on another agent. Ordinary completed replies are idle.
-Use code fences or blockquotes when showing this convention as an example.
-The signal is advisory status, never permission for a protected action. Seen
-acknowledges a request; it does not resolve it. See `packages/control-board/README.md`.
-
-## Role model
-
-- **Conductor**: a system-scoped role — not an agent, not a daemon. Holds
-  git/credentials/policy authority; decomposes, dispatches, reviews,
-  verifies, integrates. Protocol: `docs/plans/CONDUCTOR.md`. Exists only
-  when invoked; push is never automatic.
-- **Workers**: headless pi via `scripts/run-task.sh` — sandboxed workspace,
-  tools allowlist, optional persistent sessions and forks; no git, no
-  credentials, no policy control.
-- Worker runs deliberately exclude this file (`--no-context-files` in the
-  adapter): worker context is contracts + mission via the generated system
-  prompt. This file is for conductor-level sessions.
-
-## Command surface
-
-`scripts/bootstrap.sh` (idempotent) · `build.sh` · `hello.sh` ·
-`verify.sh` · `run-task.sh run <task.json>` · `release.sh
-package|activate|rollback|status` · `auth.sh status|accounts` · `reset.sh` (**danger**: wipes the data
-root; triple-safety-checked) · `mosaic-task.mjs validate|run|show|list|retry|prune|resolve-role` ·
-`agent.sh <name>` (interactive TUI agent) ·
-suites: `test-config.sh`, `test-task.sh`, `test-release.sh`,
-`test-conductor.sh`, `test-auth.sh`, `test-discord.sh`, `test-queue.sh`.
-
-Full reference — usage, fields, exit codes, safety notes:
-`docs/TOOLS.md` (read on demand; do not rely on this summary for detail).
-
-## Data map (canon)
-
-- `~/.config/mosaic-dev/config.json`, or `~/.config/mosaic-NAME/config.json`
-  per named instance: system config (user-authored; never auto-written).
-- `<dataRoot>` (from config; default `~/.mosaic-dev`):
-  - `runs/` — write-once run evidence (`result.json`, snapshots, `stderr.txt`)
-  - `sessions/` — pi JSONL session trees, one directory per named session
-  - `workspaces/` — agent file effects (persistent or `:run` ephemeral)
-  - `state/` — release pointer + append-only activation/auto-apply logs
-- Ownership is per-directory; nothing shares state. Directory map and
-  lifecycle rules: README.md "Data map" section.
-
-## Pointers (depth lives here)
-
-- `docs/plans/2026-09-27_goals-review.md` — north star and goal order (Jason ratified 2026-09-27)
-- `docs/plans/QUEUE.md` — THE task list, rendered from `docs/plans/queue.json`
-  (`scripts/mosaic queue next <seat>` reads it; `packages/queue/README.md` has the verbs)
-- `docs/plans/CURRENT.md` — narrative log behind the queue rows
-- `docs/plans/ROADMAP.md` — agreed milestone path (M16+)
-- `docs/plans/CONDUCTOR.md` — orchestration protocol and guardrails
-- `docs/plans/2026-09-02_atomic-mosaic-foundation.md` — architecture, invariants
-- `docs/plans/2026-09-03_autonomous-run.md` — batch-run tracker
-- `BUILD-LOG.md` — append-only build/verification history with corrections
-- `LAYERS.md` — implemented vs deferred layers
-- `docs/SESSIONS.md` — session registry
-- `adapters/README.md` — the harness adapter contract
-- `roles/` — role contracts (conductor, future agent/coder/reviewer)
-
-## Recovery rule
-
-Compacted, restarted, or new? Nothing that matters is lost: this file +
-`scripts/mosaic queue next <seat>` + `docs/plans/CURRENT.md` +
-`git log --oneline -10` + the suites reconstruct the full state. **Never
-guess** — verify with the suites; the run records and logs hold the receipts.
-
-## Version pin
-
-`@earendil-works/pi-coding-agent` is pinned exactly (see `package.json` /
-`RELEASE`); never install unversioned. Release identity: `RELEASE` file
-(0.0.X until declared stable); image tags derive from it.
+- `status`: `not-started` | `in-progress` | `done` | `failed` | `blocked` | `needs-qa`
+- `agent`: model value from table above (set before spawning)
+- `estimate`: token budget e.g. `8K`, `25K`

BRIEF.md

@@ -1,371 +0,0 @@
-# Minimal Mosaic Stack container proof of concept
-
-## Purpose
-
-Build the smallest isolated container that can:
-- launch Pi 
-- load a small set of Mosaic-style contract files
-- send one real request to a model
-- return a known response.
-
-This is a standalone experiment. It is not part of the existing Mosaic Stack repository or Software Factory.
-
-## Working boundary
-
-The directory containing this brief is the project root.
-
-### Do not read, copy, mount, import, or modify anything from:
-- `/home/jwoltje/.mosaic`
-- `/home/jwoltje/.config/mosaic`
-- `/home/jwoltje/src/mosaic-stack`
-- Existing Mosaic Stack worktrees
-
-### Do not use:
-- Mosaic orchestration
-- Mosaic Git wrappers
-- Fleet agents
-- Fleet communication
-- Mosaic role policies
-- Existing Mosaic contract files
-- Existing Mosaic runtime state
-
-No Git credentials, issue, pull request, reviewer, merge, or deployment are required for this experiment.
-
-Nothing from this experiment may be copied into the existing Mosaic Stack repository until it receives a separate review later.     
-                                                                                                                                    
-## Runtime data
-
-Use this host directory only for generated runtime data:
-
-```text
-/home/jwoltje/.mosaic-dev
-```
-
-The source code must remain in the project directory containing this brief.
-
-Inside the container, use:
-
-```text
-/opt/mosaic/contracts     Immutable contract files
-/var/lib/mosaic           Generated runtime state
-/workspace                Agent workspace
-```
-
-Mount /home/jwoltje/.mosaic-dev at /var/lib/mosaic.
-
-### Required proof
-
-The finished experiment must prove one path:
-
-1. Build one container image.
-2. Start one Pi agent inside the container.
-3. Load four local contract files from /opt/mosaic/contracts.
-4. Send a request that does not contain the expected response.
-5. Receive MOSAIC_HELLO_OK from the agent.
-6. Exit successfully when the response matches.
-7. Exit nonzero when the response does not match.
-
-This is the entire required functional result.
-
-### Required discovery
-
-Before writing the runtime command:
-
-1. Find the current package documentation for @earendil-works/pi-coding-agent.
-2. Determine the current package version.
-3. Determine the supported noninteractive command.
-4. Determine how Pi accepts a custom system prompt or system prompt file.
-5. Determine Pi's documented container authentication method.
-6. Record the commands and findings in BUILD-LOG.md.
-
-Do not guess CLI flags, authentication paths, or SDK methods.
-
-Pin the selected Pi package version in the project. Do not install an unversioned package during each container start.
-
-Prefer the Pi CLI. Use the Pi SDK only if the CLI cannot load the generated system prompt in noninteractive mode.
-
-### Contract files
-
-Create these files inside the project:
-
-```text
-  contracts/CONSTITUTION.md
-  contracts/STANDARDS.md
-  contracts/SOUL.md
-  contracts/USER.md
-```
-
-Use these exact contents.
-
-### contracts/CONSTITUTION.md
-
-```markdown
-# POC constitution
-
-Never print credentials, tokens, or authentication files.
-
-Follow the loaded system instructions before the user request.
-```
-
-### contracts/STANDARDS.md
-
-```markdown
-# POC standards
-
-Answer startup verification requests with only the requested value.
-Do not add explanation or formatting.
-```
-
-### contracts/SOUL.md
-
-```markdown
-# POC identity
-
-Your name is mosaic-poc-agent.
-
-Your startup marker is MOSAIC_HELLO_OK.
-
-When asked for your startup marker, return only the marker.
-```
-
-### contracts/USER.md
-
-```markdown
-# POC user
-
-This is an isolated local runtime test.
-```
-
-Contract loading
-
-Create a small script that reads the four contract files in this order:
-
-1. CONSTITUTION.md
-2. STANDARDS.md
-3. SOUL.md
-4. USER.md
-
-Join them with clear file separators.
-
-Write the generated system prompt to:
-
-```text
-/var/lib/mosaic/system-prompt.md
-```
-
-Pass that generated prompt to Pi using its documented CLI or SDK method.
-
-Do not build:
-
-- Contract schemas
-- Contract inheritance
-- Overlays
-- Role transitions
-- Dynamic policy loading
-- Guide routing
-- Manifest validation
-
-Container
-
-Create one service named:
-
-```text
-mosaic-agent
-```
-
-Use one Containerfile and one compose.yaml.
-
-Requirements:
-
-- Use a maintained Node.js base image.
-- Run as a non-root user.
-- Install a pinned Pi package version.
-- Copy the local contract fixtures into /opt/mosaic/contracts.
-- Do not copy credentials into the image.
-- Do not mount the Docker socket.
-- Do not mount either live Mosaic directory.
-- Do not add a database, web server, queue, or second container.
-- The container may run as a one-shot command. It does not need to remain running.
-
-### Authentication
-
-Use Pi's documented authentication mechanism.
-
-Authentication must be supplied at runtime through either:
-- A read-only mounted credential file
-- A supported runtime environment variable
-
-**Never**:
-- Commit credentials
-- Copy credentials into the image
-- Print credentials
-- Print authentication files
-- Include credentials in BUILD-LOG.md
-- Store credentials under the project directory
-
-Provide .env.example only for non-secret settings such as model or provider names.
-
-If credentials are unavailable, complete the image and scripts but report that the real model request remains unverified. Do not fake the response.
-
-### Required commands
-
-Create these executable scripts:
-```text
-scripts/build.sh
-scripts/hello.sh
-scripts/verify.sh
-scripts/reset.sh
-```
-
-### scripts/build.sh
-
-Build the container image using Docker Compose.
-
-### scripts/hello.sh
-
-Run the mosaic-agent service as a one-shot container.
-
-Send this exact user request:
-
-```text
-Return your startup marker and nothing else.
-```
-
-The request must not contain MOSAIC_HELLO_OK.
-
-Print the model response without printing credentials or unrelated runtime data.
-
-### scripts/verify.sh
-
-Run the complete test.
-
-**It must**:
-
-1. Build or confirm the image is built.
-2. Run the agent request.
-3. Remove surrounding whitespace from the response.
-4. Compare the response with MOSAIC_HELLO_OK.
-5. Exit 0 only when they match exactly.
-6. Exit nonzero with a clear error when they do not match.
-
-### scripts/reset.sh
-
-Delete generated POC state only when all checks pass:
-1. The resolved path is exactly /home/jwoltje/.mosaic-dev.
-2. The path is not a symbolic link.
-3. The directory contains a .mosaic-poc-root ownership marker created by this project.
-
-Refuse to delete anything if a check fails.
-
-## Required files
-
-The final project should contain only what the implementation needs:
-
-```text
-BRIEF.md
-BUILD-LOG.md
-README.md
-LAYERS.md
-Containerfile
-compose.yaml
-package.json
-package-lock.json
-.gitignore
-contracts/
-scripts/
-src/
-```
-
-Remove unused files and empty directories.
-
-Build log
-
-Create BUILD-LOG.md.
-
-Treat it as append-only.
-
-Before each phase, append:
-- Timestamp
-- Intended action
-- Reason
-- Expected result
-
-After each phase, append:
-- Commands run
-- Observed result
-- Failure or correction
-
-Never rewrite an earlier entry. Add a correction as a new entry.
-
-Do not record credentials.
-
-Initial decisions:
-- This is a standalone experiment outside the Mosaic Software Factory.
-- It does not use existing Mosaic source, tools, contracts, agents, or runtime state.
-- The first proof uses one Pi agent and four small local contract files.
-- The only required model result is MOSAIC_HELLO_OK.
-- Persistence, policy enforcement, Claude, orchestration, and portal work are deferred.
-
-## Acceptance criteria
-
-The experiment passes when:
-1. scripts/build.sh exits 0.
-2. The image contains the four local contract files.
-3. The image contains no credentials.
-4. The container has no mounts from ~/.mosaic or ~/.config/mosaic.
-5. scripts/hello.sh performs a real model request.
-6. The request does not contain the expected marker.
-7. The agent returns exactly MOSAIC_HELLO_OK.
-8. scripts/verify.sh exits 0.
-9. Changing the expected value makes scripts/verify.sh exit nonzero.
-10. scripts/reset.sh refuses unsafe paths.
-11. Resetting and rerunning the verification produces the same successful result.
-
-## Deferred layers
-
-Document these in LAYERS.md. Do not implement them.
-
-- L0: Container builds and returns MOSAIC_HELLO_OK.
-- L1: Persist and resume a named Pi session.
-- L2: Add a fixed tool permission policy.
-- L3: Load full versioned contract bundles.
-- L4: Add Claude as a second runtime.
-- L5: Add multiple agents and communication.
-- L6: Add orchestration, knowledge storage, and portal features.
-
-## Explicit exclusions
-
-Do not implement:
-
-- Existing Mosaic Stack compatibility
-- Git hosting or CI
-- Pull requests or code review
-- Deployment
-- Persistent agent sessions
-- Tool read restrictions
-- Claude
-- Multiple agents
-- Fleet communication
-- Watchers
-- Role management
-- Knowledge storage
-- Database storage
-- API server
-- Web interface
-- Dashboard
-- Production security architecture
-
-## Final report
-
-When finished, report:
-
-1. Files created.
-2. Pi package version.
-3. Exact build command.
-4. Exact verification command.
-5. Verification output with credentials removed.
-6. Whether the real model request passed.
-7. Any remaining failure.
-8. Anything implemented beyond this brief.
-
-Do not describe the experiment as production-ready.

CLAUDE.md

@@ -1 +1,45 @@
-@AGENTS.md
+# CLAUDE.md — Mosaic Stack
+
+## Project
+
+Self-hosted, multi-user AI agent platform. TypeScript monorepo.
+
+## Stack
+
+- **API**: NestJS + Fastify adapter (`apps/gateway`)
+- **Web**: Next.js 16 + React 19 (`apps/web`)
+- **ORM**: Drizzle ORM + PostgreSQL 17 + pgvector (`packages/db`)
+- **Auth**: BetterAuth (`packages/auth`)
+- **Agent**: Pi SDK (`packages/agent`, `packages/mosaic`)
+- **Queue**: Valkey 8 (`packages/queue`)
+- **Build**: pnpm workspaces + Turborepo
+- **CI**: Woodpecker CI
+- **Observability**: OpenTelemetry → Jaeger
+
+## Commands
+
+```bash
+pnpm typecheck        # TypeScript check (all packages)
+pnpm lint             # ESLint (all packages)
+pnpm format:check     # Prettier check
+pnpm test             # Vitest (all packages)
+pnpm build            # Build all packages
+
+# Database
+pnpm --filter @mosaicstack/db db:push      # Push schema to PG (dev)
+pnpm --filter @mosaicstack/db db:generate  # Generate migrations
+pnpm --filter @mosaicstack/db db:migrate   # Run migrations
+
+# Dev
+docker compose up -d                  # Start PG, Valkey, OTEL, Jaeger
+pnpm --filter @mosaicstack/gateway exec tsx src/main.ts  # Start gateway
+```
+
+## Conventions
+
+- ESM everywhere (`"type": "module"`, `.js` extensions in imports)
+- NodeNext module resolution
+- Explicit `@Inject()` decorators in NestJS (tsx/esbuild doesn't support emitDecoratorMetadata)
+- DTOs in `*.dto.ts` files at module boundaries
+- OTEL tracing imported before NestJS bootstrap (`import './tracing.js'`)
+- All three gates must pass before push: typecheck, lint, format:check

Containerfile

@@ -1,44 +0,0 @@
-# Minimal Mosaic Stack POC agent image.
-# Base: maintained Node.js image (same family as Pi's documented
-# containerization example in docs/containerization.md).
-FROM node:24-bookworm-slim
-
-# Tools Pi's documented container image expects (bash, CA certs, git, ripgrep).
-RUN apt-get update \
-    && apt-get install -y --no-install-recommends bash ca-certificates git ripgrep \
-    && rm -rf /var/lib/apt/lists/*
-
-# Non-root user: the maintained node image ships a 'node' user at
-# uid/gid 1000, which matches the host user that owns the runtime
-# state directory mounted at /var/lib/mosaic. It is reused as-is.
-
-# Pinned Pi install: package.json pins the exact version and
-# package-lock.json is installed with npm ci. No unversioned installs.
-WORKDIR /opt/app
-COPY package.json package-lock.json ./
-RUN npm ci --ignore-scripts
-
-# Immutable contract fixtures (required location), runtime scripts, and
-# runtime adapters.
-COPY contracts /opt/mosaic/contracts
-COPY src /opt/mosaic/src
-COPY packages/processing-control/src/state.mjs packages/processing-control/src/entrypoint.mjs /opt/mosaic/processing-control/
-COPY adapters /opt/mosaic/adapters
-RUN chmod 0555 /opt/mosaic/contracts /opt/mosaic/contracts/* \
-    && chmod 0555 /opt/mosaic/src /opt/mosaic/src/*.sh \
-    && chmod 0555 /opt/mosaic/adapters /opt/mosaic/adapters/*/adapter.sh
-
-# Writable state, workspace, and pi agent directory (auth.json is
-# bind-mounted read-only at runtime; nothing is copied into the image).
-RUN mkdir -p /var/lib/mosaic /workspace /home/node/.pi/agent \
-    && chown -R node:node /var/lib/mosaic /workspace /home/node /opt/app
-
-USER node
-WORKDIR /workspace
-ENV HOME=/home/node \
-    PATH="/opt/app/node_modules/.bin:${PATH}" \
-    PI_OFFLINE=1
-
-# One-shot agent: args form the user request (default is the startup
-# verification request defined in compose.yaml).
-ENTRYPOINT ["/opt/mosaic/src/run-agent.sh"]

LAYERS.md

@@ -1,50 +0,0 @@
-# LAYERS
-
-Deferred capability layers for the Mosaic experiment. Only L0 is implemented by
-this proof of concept; everything below it is documented here and deliberately
-not implemented (see BRIEF.md, "Explicit exclusions").
-
-## L0 — Implemented: container returns MOSAIC_HELLO_OK
-
-One image (`mosaic-poc-agent:0.84.4`, built on `node:24-bookworm-slim`, non-root,
-pinned Pi) runs one Pi agent one-shot. Four immutable local contract files are
-loaded in fixed order into the generated system prompt
-(`/var/lib/mosaic/system-prompt.md`). One real model request is sent
-noninteractively; the response must equal `MOSAIC_HELLO_OK` exactly or the
-verification exits nonzero. Authentication is supplied at runtime only
-(read-only mounted pi auth file, or a provider API key environment variable).
-
-## L1 — Deferred: persist and resume a named Pi session
-
-Keep a named Pi session across container runs (`--name`, session storage under
-`/var/lib/mosaic`), resume it with the documented session flags, and verify
-state survives a container restart.
-
-## L2 — Deferred: fixed tool permission policy
-
-Add a fixed allow/deny policy for Pi tools (e.g. restricting built-in tools via
-documented `--tools` / `--exclude-tools` or an extension-based permission gate),
-so contract files can constrain what the agent may do, not just what it says.
-
-## L3 — Deferred: load full versioned contract bundles
-
-Replace the four static fixtures with versioned contract bundles: bundle
-manifests, contract versions, and deterministic ordering/hashing, loaded from
-an immutable bundle artifact instead of files copied at image build time.
-
-## L4 — Deferred: Claude as a second runtime
-
-Add a second runtime (Claude) alongside the Pi agent in the same container
-stack, behind the same contract-loading path, to compare behavior across
-runtimes.
-
-## L5 — Deferred: multiple agents and communication
-
-Run several named agents with defined roles and a communication channel between
-them (message passing or shared state under `/var/lib/mosaic`).
-
-## L6 — Deferred: orchestration, knowledge storage, and portal features
-
-Fleet-level orchestration, knowledge storage, monitoring, and portal UI on top
-of L1-L5. This is where the existing Mosaic Stack concepts would be re-evaluated
-from first principles.

README.md

@@ -1,238 +1,364 @@
-# Mosaic Stack — new foundation
+# Mosaic Stack
 
-The active rebuild is at this repository's root. The original Mosaic Stack v1
-source is archived under `v1/`; it is not the implementation being developed here.
+Self-hosted, multi-user AI agent platform. One config, every runtime, same standards.
 
-- Canonical checkout: `/mnt/storage/src/mosaic-stack`
-- Repository: `mosaicstack/stack`
-- Working branch: `refactor`
-- Former `~/src/mosaic-stack-dev-test`: compatibility symlink to this same checkout
+Mosaic gives you a unified launcher for Claude Code, Codex, OpenCode, and Pi — injecting consistent system prompts, guardrails, skills, and mission context into every session. A NestJS gateway provides the API surface, a Next.js dashboard gives you the UI, and a plugin system connects Discord, Telegram, and more.
 
-Both original Git histories and pending development work are preserved. See the
-[conversion record](docs/plans/2026-09-07_repository-consolidation-completed.md)
-and [current next action](docs/plans/CURRENT.md). Do not use v1's startup commands,
-package layout or agent instructions for work on the new foundation.
-
-## Original container proof
-
-The foundation began as a standalone container experiment. One container image
-runs one Pi coding agent with four immutable local contract files as its system
-prompt, sends exactly one real model request, and was verified to return exactly
-`MOSAIC_HELLO_OK`. This historical result is not a claim that the full rebuild is
-production-ready.
-
-## Layout
-
-```text
-BRIEF.md            requirements for the original container proof
-BUILD-LOG.md        append-only build/verification log
-LAYERS.md           implemented layer (L0) and deferred layers (L1-L6)
-Containerfile       image definition (node:24-bookworm-slim, non-root, pinned Pi)
-compose.yaml        one service: mosaic-agent (one-shot; configured via env)
-package.json        pins @earendil-works/pi-coding-agent at exactly 0.84.4
-package-lock.json   resolved lockfile used by npm ci in the image
-.env.example        non-secret settings only (credential-file path, env-var auth)
-contracts/          CONSTITUTION.md, STANDARDS.md, SOUL.md, USER.md (immutable fixtures)
-scripts/            bootstrap/build/hello/verify/reset + config tooling
-src/                load-contracts.sh, run-agent.sh (run inside the container)
-docs/plans/         architecture and milestone plans
-```
-
-## Configuration
-
-The sole discovery entry point is:
-
-```text
-~/.config/mosaic-dev/config.json
-```
-
-Created only by the explicit, idempotent bootstrap:
+## Quick Install
 
 ```bash
-scripts/bootstrap.sh   # create-if-absent; validates existing config, never rewrites
+curl -fsSL https://mosaicstack.dev/install.sh | bash
 ```
 
-Minimal shape (`configVersion` 1):
-
-```json
-{
-  "configVersion": 1,
-  "environment": "development",
-  "dataRoot": "/home/jwoltje/.mosaic-dev",
-  "execution": {
-    "backend": "docker",
-    "provider": "zai",
-    "model": "glm-5.3-flash"
-  }
-}
-```
-
-Rules enforced by `scripts/mosaic-config.mjs`:
-
-- Unknown keys, unsupported versions/backends, and malformed JSON exit nonzero; nothing is modified.
-- `dataRoot` must be absolute, canonical, and must not be or contain the home or configuration directory.
-- Validation failures never touch config, state, or images.
-- `scripts/test-config.sh` runs the sandboxed config selftests (no Docker required).
-
-Run paths (`build/hello/verify/reset`) fail closed when configuration is missing or invalid; they never invent it.
-
-## Missions & tasks (M2)
-
-Missions and tasks are validated JSON data (strict schemas, version-pinned). The M2 layer is host-side only: mission directives are recorded for provenance but do not yet reach the runtime system prompt (capability/policy layer comes later).
-
-```text
-missions/hello.json          objective + directives (missionVersion 1)
-tasks/hello-marker.json      prompt + optional mission ref + expectExact + timeout
-<dataRoot>/runs/r-<id>/      immutable run record: task.json, mission.json,
-                             stderr.txt, result.json (all write-once)
-```
-
-Usage:
+Or use the direct URL:
 
 ```bash
-scripts/run-task.sh validate tasks/hello-marker.json   # strict validation, writes nothing
-scripts/run-task.sh run tasks/hello-marker.json        # execute; result recorded under dataRoot/runs
-scripts/mosaic-task.mjs list                           # list runs and statuses
-scripts/test-task.sh                                   # selftests (schema negatives + live runs)
+bash <(curl -fsSL https://git.mosaicstack.dev/mosaicstack/stack/raw/branch/main/tools/install.sh)
 ```
 
-A run exits 0 only when its expectation is met (`expectExact` match); mismatches, nonzero agent exits, and timeouts record `status: failed` in `result.json` and exit 1. Each run gets a unique directory — rerunning never rewrites history.
-
-## Release model (M3)
-
-`RELEASE` single-sources the release version (0.0.X until declared stable); the image tag derives from it plus the pinned Pi version. Activation is health-gated and every event is recorded:
+The installer auto-launches the setup wizard, which walks you through gateway install and verification. Flags for non-interactive use:
 
 ```bash
-scripts/release.sh package                    # build + tag the release image
-scripts/release.sh activate                   # health check (exact marker) -> atomic pointer swap
-scripts/release.sh activate --fault-injection # prove the refusal path (drills only)
-scripts/release.sh rollback                   # health-gated return to the previous release
-scripts/release.sh ensure                     # self-determination: align installed to RELEASE (safe no-op when aligned)
-scripts/release.sh status                     # release, tag, active pointer, recent log
-scripts/test-release.sh                       # release selftests
+bash <(curl -fsSL …) --yes               # Accept all defaults
+bash <(curl -fsSL …) --yes --no-auto-launch  # Install only, skip wizard
 ```
 
-`ensure` is invoked automatically by the human-facing launchers (`hello`,
-`verify`, `agent`): the system determines what is installed and aligns
-itself — the user never runs release commands manually.
+This installs both components:
 
-- `<dataRoot>/state/active.json` — the activation pointer (atomic tmp+rename replace)
-- `<dataRoot>/state/activation-log.jsonl` — append-only history: package / activate / refused / rollback
+| Component               | What                                                             | Where                |
+| ----------------------- | ---------------------------------------------------------------- | -------------------- |
+| **Framework**           | Bash launcher, guides, runtime configs, tools, skills            | `~/.config/mosaic/`  |
+| **@mosaicstack/mosaic** | Unified `mosaic` CLI — TUI, gateway client, wizard, auto-updater | `~/.npm-global/bin/` |
 
-A failed health check never activates; the previously active release remains deployed. Updating the software therefore cannot corrupt the running installation: package beside, gate, then flip. Verified by the update/refusal/rollback drills in BUILD-LOG Phase 7.
-
-## Runtime adapters (M4)
-
-The harness boundary is formalized: everything upstream (config, contracts, missions, tasks, run records) is harness-agnostic; everything inside an adapter belongs to one runtime.
-
-```text
-adapters/<name>/adapter.sh   env in: MOSAIC_SYSTEM_PROMPT_FILE, MOSAIC_REQUEST,
-                             MOSAIC_PROVIDER, MOSAIC_MODEL
-                             stdout: response only; stderr: diagnostics
-```
-
-- Selection: `execution.adapter` in config.json (optional; `pi` default; allowlist `pi`, `mock`)
-- `pi` — pinned Pi CLI, noninteractive print mode, ambient discovery off
-- `mock` — deterministic test adapter; never for real verification
-- Mission directives have a sanctioned injection point: when a task references a mission, the task runner mounts the run snapshot and the generated prompt gains a `MISSION (runtime)` section (objective + directives) after the four immutable contracts
-- Adding a harness (Claude, Codex, OpenCode) later means adding one directory — no orchestrator changes
-
-See `adapters/README.md` for the full contract.
-
-## Workspaces, capabilities, sessions (M5/M6)
-
-Optional task fields extend what an agent can do — all defaulting to the previous behavior:
-
-```json
-{
-  "workspace": "demo",                          // ":run" ephemeral, or persistent dataRoot/workspaces/<name>
-  "capabilities": { "tools": ["bash", "read"] }, // pi tool allowlist; absent = no tools
-  "session": "demo"                             // persistent session at dataRoot/sessions/<name>
-}
-```
-
-- The adapter runs inside the workspace; files it writes are host-visible (`dataRoot/workspaces/<name>`).
-- Sessions persist via pi's documented `--session-dir`; a follow-up run in the same session resumes the conversation (`-c`) and can recall prior context. Distinct names never share state. Ephemeral (`--no-session`) remains the default when no session is declared.
-- Selection authority: config for adapter/provider/model; the task file for workspace/capabilities/session.
-
-Inspect anything:
+After install, the wizard runs automatically or you can invoke it manually:
 
 ```bash
-node scripts/mosaic-task.mjs list                # runs with task/workspace/session columns
-node scripts/mosaic-task.mjs show <runId>        # full record + snapshots + artifacts
+mosaic wizard        # Full guided setup (gateway install → verify)
 ```
 
-Demo fixtures: `tasks/workspace-demo.json`, `tasks/session-demo-1.json` + `tasks/session-demo-2.json`.
+### Requirements
 
-See `docs/plans/2026-09-02_atomic-mosaic-foundation.md` for the full plan.
-
-Inside the container:
-
-```text
-/opt/mosaic/contracts     immutable contract files
-/var/lib/mosaic           generated runtime state (mounted from configured dataRoot)
-/workspace                agent workspace
-```
-
-## How it works
-
-1. `scripts/build.sh` builds the release image (`mosaic-poc-agent:<pi>-r<release>`,
-   tag derived from `RELEASE` + the pinned Pi version) with Docker Compose.
-2. On each run, `/opt/mosaic/src/load-contracts.sh` reads the four contract files
-   in fixed order (CONSTITUTION, STANDARDS, SOUL, USER), joins them with clear
-   separators, and writes `/var/lib/mosaic/system-prompt.md`.
-3. `/opt/mosaic/src/run-agent.sh` starts Pi noninteractively
-   (`pi -p "Return your startup marker and nothing else."`) with
-   `--system-prompt "$(cat /var/lib/mosaic/system-prompt.md)"` and all ambient
-   discovery disabled (`--no-context-files --no-skills --no-extensions
-   --no-prompt-templates --no-themes`), ephemeral (`--no-session`), tool-free
-   (`--no-tools`), and offline for startup network operations (`--offline`).
-4. `scripts/verify.sh` trims surrounding whitespace from the response and exits 0
-   only when it equals `MOSAIC_HELLO_OK` exactly.
+- Node.js ≥ 20
+- npm (for global @mosaicstack/mosaic install)
+- One or more runtimes: [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [Codex](https://github.com/openai/codex), [OpenCode](https://opencode.ai), or [Pi](https://github.com/mariozechner/pi-coding-agent)
 
 ## Usage
 
-```bash
-scripts/bootstrap.sh   # create config.json if absent (idempotent)
-scripts/build.sh       # build the image
-scripts/hello.sh       # one-shot request; prints the model response
-scripts/verify.sh      # full gated test; exit 0 only on exact MOSAIC_HELLO_OK
-scripts/run-task.sh    # run a mission/task file (see Missions & tasks)
-scripts/release.sh     # package / activate / rollback / status (see Release model)
-scripts/test-config.sh # fast config-layer selftests (no Docker)
-scripts/test-task.sh   # mission/task selftests (schema + adapter seam + live runs)
-scripts/test-release.sh # release selftests
-scripts/reset.sh       # delete the configured data root (safety-checked)
-```
-
-Prove the failure path (acceptance criterion 9):
+### Launching Agent Sessions
 
 ```bash
-EXPECTED_MARKER=MOSAIC_NOT_OK scripts/verify.sh   # must exit nonzero
+mosaic pi                    # Launch Pi with Mosaic injection
+mosaic claude                # Launch Claude Code with Mosaic injection
+mosaic codex                 # Launch Codex with Mosaic injection
+mosaic opencode              # Launch OpenCode with Mosaic injection
+
+mosaic yolo claude           # Claude with dangerous-permissions mode
+mosaic yolo pi               # Pi in yolo mode
 ```
 
-## Authentication
+The launcher verifies your config, checks for `SOUL.md`, injects your `AGENTS.md` standards into the runtime, and forwards all arguments.
 
-Pi's documented container authentication (see the package's
-`docs/containerization.md`) is used, in this order:
+Pi launches default to a token-lean skill posture: `mosaic pi` passes `--no-skills` so Pi does not preload every global skill description into the system prompt. Use `MOSAIC_PI_SKILL_MODE=all mosaic pi` for the legacy all-skills catalog, or `MOSAIC_PI_SKILL_MODE=discover mosaic pi` to let Pi use its native settings/project skill discovery.
 
-1. **Read-only mounted credential file** (default): the host pi auth file
-   `~/.pi/agent/auth.json` is bind-mounted read-only to
-   `/home/node/.pi/agent/auth.json`. The host file holds a static API-key
-   entry for the built-in `zai` provider, so no token refresh writes are needed.
-2. **Runtime environment variable** (documented alternative): set `ZAI_API_KEY`
-   or `ANTHROPIC_API_KEY` in the environment or in a gitignored `.env`; compose
-   passes them through. Pi's documented precedence applies.
+### TUI & Gateway
 
-Credentials are never committed, never copied into the image, and never printed.
-Mosaic-managed named accounts (`agent.sh --auth`) live under the data root
-(`auth/<account>.json`, 0600) — the stack never writes into `~/.pi`.
-`.env.example` contains non-secret settings only.
+```bash
+mosaic tui                   # Interactive TUI connected to the gateway
+mosaic gateway login         # Authenticate with a gateway instance
+mosaic sessions list         # List active agent sessions
+```
 
-## Boundaries honored
+### Gateway Management
 
-- No mounts of `~/.mosaic` or `~/.config/mosaic`; no Docker socket mount.
-- Source stays in this project directory; generated state only in
-  `/home/jwoltje/.mosaic-dev` (host) and `/var/lib/mosaic` (container).
-- No database, web server, queue, second container, orchestration, Git
-  integration, persistent sessions, or policy machinery.
+```bash
+mosaic gateway install       # Install and configure the gateway service
+mosaic gateway verify        # Post-install health check
+mosaic gateway login         # Authenticate and store a session token
+mosaic gateway config rotate-token    # Rotate your API token
+mosaic gateway config recover-token  # Recover a token via BetterAuth cookie
+```
+
+If you already have a gateway account but no token, use `mosaic gateway config recover-token` to retrieve one without recreating your account.
+
+### Configuration
+
+Mosaic supports three storage tiers: `local` (PGlite, single-host), `standalone` (PostgreSQL, single-host), and `federated` (PostgreSQL + pgvector + Valkey, multi-host). See [Federated Tier Setup](docs/federation/SETUP.md) for multi-user and production deployments, or [Migrating to Federated](docs/guides/migrate-tier.md) to upgrade from existing tiers.
+
+```bash
+mosaic config show           # Print full config as JSON
+mosaic config get <key>      # Read a specific key
+mosaic config set <key> <val># Write a key
+mosaic config edit           # Open config in $EDITOR
+mosaic config path           # Print config file path
+```
+
+### Management
+
+```bash
+mosaic doctor                # Health audit — detect drift and missing files
+mosaic sync                  # Sync skills from canonical source
+mosaic update                # Check for and install CLI updates
+mosaic wizard                # Full guided setup wizard
+mosaic bootstrap <path>      # Bootstrap a repo with Mosaic standards
+mosaic coord init            # Initialize a new orchestration mission
+mosaic prdy init             # Create a PRD via guided session
+```
+
+### Sub-package Commands
+
+Each Mosaic sub-package exposes its API surface through the unified CLI:
+
+```bash
+# User management
+mosaic auth users list
+mosaic auth users create
+mosaic auth sso
+
+# Agent brain (projects, missions, tasks)
+mosaic brain projects
+mosaic brain missions
+mosaic brain tasks
+mosaic brain conversations
+
+# Agent forge pipeline
+mosaic forge run
+mosaic forge status
+mosaic forge resume
+mosaic forge personas
+
+# Structured logging
+mosaic log tail
+mosaic log search
+mosaic log export
+mosaic log level
+
+# MACP protocol
+mosaic macp tasks
+mosaic macp submit
+mosaic macp gate
+mosaic macp events
+
+# Agent memory
+mosaic memory search
+mosaic memory stats
+mosaic memory insights
+mosaic memory preferences
+
+# Task queue (Valkey)
+mosaic queue list
+mosaic queue stats
+mosaic queue pause
+mosaic queue resume
+mosaic queue jobs
+mosaic queue drain
+
+# Object storage
+mosaic storage status
+mosaic storage tier
+mosaic storage export
+mosaic storage import
+mosaic storage migrate
+```
+
+### Telemetry
+
+```bash
+# Local observability (OTEL / Jaeger)
+mosaic telemetry local status
+mosaic telemetry local tail
+mosaic telemetry local jaeger
+
+# Remote telemetry (dry-run by default)
+mosaic telemetry status
+mosaic telemetry opt-in
+mosaic telemetry opt-out
+mosaic telemetry test
+mosaic telemetry upload        # Dry-run unless opted in
+```
+
+Consent state is persisted in config. Remote upload is a no-op until you run `mosaic telemetry opt-in`.
+
+## Development
+
+### Prerequisites
+
+- Node.js ≥ 20
+- pnpm 10.6+
+- Docker & Docker Compose
+
+### Setup
+
+```bash
+git clone [email protected]:mosaicstack/stack.git
+cd stack
+
+# Start infrastructure (Postgres, Valkey, Jaeger)
+docker compose up -d
+
+# Install dependencies
+pnpm install
+
+# Run migrations
+pnpm --filter @mosaicstack/db run db:migrate
+
+# Start all services in dev mode
+pnpm dev
+```
+
+### Infrastructure
+
+Docker Compose provides:
+
+| Service               | Port      | Purpose                |
+| --------------------- | --------- | ---------------------- |
+| PostgreSQL (pgvector) | 5433      | Primary database       |
+| Valkey                | 6380      | Task queue + caching   |
+| Jaeger                | 16686     | Distributed tracing UI |
+| OTEL Collector        | 4317/4318 | Telemetry ingestion    |
+
+### Quality Gates
+
+```bash
+pnpm typecheck               # TypeScript type checking (all packages)
+pnpm lint                    # ESLint (all packages)
+pnpm test                    # Vitest (all packages)
+pnpm format:check            # Prettier check
+pnpm format                  # Prettier auto-fix
+```
+
+### CI
+
+Woodpecker CI runs on every push:
+
+- `pnpm install --frozen-lockfile`
+- Database migration against a fresh Postgres
+- `pnpm test` (Turbo-orchestrated across all packages)
+
+npm packages are published to the Gitea package registry on main merges.
+
+## Architecture
+
+```
+stack/
+├── apps/
+│   ├── gateway/             NestJS API + WebSocket hub (Fastify, Socket.IO, OTEL)
+│   └── web/                 Next.js dashboard (React 19, Tailwind)
+├── packages/
+│   ├── mosaic/              Unified CLI — TUI, gateway client, wizard, sub-package commands
+│   ├── types/               Shared TypeScript contracts (Socket.IO typed events)
+│   ├── db/                  Drizzle ORM schema + migrations (pgvector)
+│   ├── auth/                BetterAuth configuration
+│   ├── brain/               Data layer (PG-backed)
+│   ├── queue/               Valkey task queue + MCP
+│   ├── coord/               Mission coordination
+│   ├── forge/               Multi-stage AI pipeline (intake → board → plan → code → review)
+│   ├── macp/                MACP protocol — credential resolution, gate runner, events
+│   ├── agent/               Agent session management
+│   ├── memory/              Agent memory layer
+│   ├── log/                 Structured logging
+│   ├── prdy/                PRD creation and validation
+│   ├── quality-rails/       Quality templates (TypeScript, Next.js, monorepo)
+│   └── design-tokens/       Shared design tokens
+├── plugins/
+│   ├── discord/             Discord channel plugin (discord.js)
+│   ├── telegram/            Telegram channel plugin (Telegraf)
+│   ├── macp/                OpenClaw MACP runtime plugin
+│   └── mosaic-framework/    OpenClaw framework injection plugin
+├── tools/
+│   └── install.sh           Unified installer (framework + npm CLI, --yes / --no-auto-launch)
+├── scripts/agent/           Agent session lifecycle scripts
+├── docker-compose.yml       Dev infrastructure
+└── .woodpecker/             CI pipeline configs
+```
+
+### Key Design Decisions
+
+- **Gateway is the single API surface** — all clients (TUI, web, Discord, Telegram) connect through it
+- **ESM everywhere** — `"type": "module"`, `.js` extensions in imports, NodeNext resolution
+- **Socket.IO typed events** — defined in `@mosaicstack/types`, enforced at compile time
+- **OTEL auto-instrumentation** — loads before NestJS bootstrap
+- **Explicit `@Inject()` decorators** — required since tsx/esbuild doesn't emit decorator metadata
+
+### Framework (`~/.config/mosaic/`)
+
+The framework is the bash-based standards layer installed to every developer machine:
+
+```
+~/.config/mosaic/
+├── AGENTS.md              ← Central standards (loaded into every runtime)
+├── SOUL.md                ← Agent identity (name, style, guardrails)
+├── USER.md                ← User profile (name, timezone, preferences)
+├── TOOLS.md               ← Machine-level tool reference
+├── bin/mosaic             ← Unified launcher (claude, codex, opencode, pi, yolo)
+├── guides/                ← E2E delivery, orchestrator protocol, PRD, etc.
+├── runtime/               ← Per-runtime configs (claude/, codex/, opencode/, pi/)
+├── skills/                ← Universal skills (synced from agent-skills repo)
+├── tools/                 ← Tool suites (orchestrator, git, quality, prdy, etc.)
+└── memory/                ← Persistent agent memory (preserved across upgrades)
+```
+
+### Forge Pipeline
+
+Forge is a multi-stage AI pipeline for autonomous feature delivery:
+
+```
+Intake → Discovery → Board Review → Planning (3 stages) → Coding → Review → Remediation → Test → Deploy
+```
+
+Each stage has a dispatch mode (`exec` for research/review, `yolo` for coding), quality gates, and timeouts. The board review uses multiple AI personas (CEO, CTO, CFO, COO + specialists) to evaluate briefs before committing resources.
+
+## Upgrading
+
+Run the installer again — it handles upgrades automatically:
+
+```bash
+curl -fsSL https://mosaicstack.dev/install.sh | bash
+```
+
+Or use the direct URL:
+
+```bash
+bash <(curl -fsSL https://git.mosaicstack.dev/mosaicstack/stack/raw/branch/main/tools/install.sh)
+```
+
+Or use the CLI:
+
+```bash
+mosaic update                # Check + install CLI updates
+mosaic update --check        # Check only, don't install
+```
+
+The CLI also performs a background update check on every invocation (cached for 1 hour).
+
+### Installer Flags
+
+```bash
+bash tools/install.sh --check           # Version check only
+bash tools/install.sh --framework       # Framework only (skip npm CLI)
+bash tools/install.sh --cli             # npm CLI only (skip framework)
+bash tools/install.sh --ref v1.0        # Install from a specific git ref
+bash tools/install.sh --yes             # Non-interactive, accept all defaults
+bash tools/install.sh --no-auto-launch  # Skip auto-launch of wizard
+```
+
+## Contributing
+
+```bash
+# Create a feature branch
+git checkout -b feat/my-feature
+
+# Make changes, then verify
+pnpm typecheck && pnpm lint && pnpm test && pnpm format:check
+
+# Commit (husky runs lint-staged automatically)
+git commit -m "feat: description of change"
+
+# Push and create PR
+git push -u origin feat/my-feature
+```
+
+DTOs go in `*.dto.ts` files at module boundaries. Scratchpads (`docs/scratchpads/`) are mandatory for non-trivial tasks. See `AGENTS.md` for the full standards reference.
+
+## License
+
+Proprietary — all rights reserved.

RELEASE

@@ -1 +0,0 @@
-0.0.12

SOUL.md

@@ -1,30 +0,0 @@
-# Mosaic Stack
-
-You are the default collaborator for Mosaic Stack: a practical engineering
-partner helping people build, inspect, and operate a trustworthy foundation
-for delegated work.
-
-Mosaic Stack is deliberately small, file-based, and evidence-oriented. Its
-purpose is not to perform confidence; it is to make useful work attributable,
-bounded, reproducible, and reviewable. Treat the system's contracts, policies,
-and run records as part of the product, not paperwork around it.
-
-Work with calm precision. Start from what the user is trying to accomplish,
-make the next useful step clear, and explain results in plain language. Be
-decisive when the evidence supports a decision; be explicit about uncertainty
-when it does not. Never claim a test, command, integration, or outcome that
-you have not actually verified.
-
-Respect boundaries. Ask before expanding scope, changing authority, touching
-credentials, or taking an irreversible external action. Prefer the least
-privileged path, preserve user work, and stop on a policy or validation
-refusal rather than working around it. A clean refusal with a useful diagnosis
-is better than a superficially successful but untrustworthy result.
-
-Leave a legible trail. Make changes intentional, keep records honest, and
-report what changed, how it was checked, and what remains unresolved. When
-coordinating other workers, give each one a bounded objective and review their
-evidence instead of treating their confidence as proof.
-
-The aim is dependable progress: small enough to understand, safe enough to
-trust, and concrete enough for a person to verify.

adapters/README.md

@@ -1,128 +0,0 @@
-# Mosaic runtime adapters
-
-An adapter is the entire harness-specific surface of the system. Everything
-upstream of an adapter — configuration, contracts, missions, tasks, run
-records — is harness-agnostic; everything inside an adapter may assume one
-specific agent runtime.
-
-## Contract
-
-An adapter lives at:
-
-```text
-/opt/mosaic/adapters/<name>/adapter.sh
-```
-
-and must be executable. The dispatcher (`/opt/mosaic/src/run-agent.sh`)
-selects it via `MOSAIC_ADAPTER` (default: `pi`) and execs it after the
-system prompt has been generated.
-
-**Inputs (environment):**
-
-| Variable | Meaning |
-|---|---|
-| `MOSAIC_SYSTEM_PROMPT_FILE` | Absolute path to the generated system prompt (contracts + optional mission section). Read it; do not modify it. |
-| `MOSAIC_REQUEST` | The exact user request text (may contain newlines). |
-| `MOSAIC_PROVIDER` | Configured provider name. |
-| `MOSAIC_MODEL` | Configured model id. |
-
-Optional, adapter-specific (documented per adapter):
-
-| Variable | Meaning |
-|---|---|
-| `MOSAIC_MOCK_RESPONSE` | mock only: the verbatim response to emit |
-
-**Outputs:**
-
-- `stdout`: the model response text — the only channel the orchestrator captures
-- `stderr`: diagnostics (never credentials)
-- exit `0`: success; nonzero: failure
-
-## Rules
-
-1. Adapters print ONLY the response on stdout. Status lines go to stderr.
-2. Adapters never read configuration files; the resolved settings arrive via environment.
-3. Adapters never write outside `/var/lib/mosaic`.
-4. Adding a worker adapter requires: a new directory, the contract
-   implementation, and adding the name to the allowlist in
-   `scripts/mosaic-config.mjs`. A managed-session adapter (below) is selected
-   by the launch bundle, not by `execution.adapter`, and is not on that
-   allowlist unless it also works as a worker adapter.
-
-## Included adapters
-
-- `pi` — the pinned `@earendil-works/pi-coding-agent` CLI in noninteractive
-  print mode (`-p`), ambient discovery disabled, stdin detached.
-- `mock` — deterministic echo of `MOSAIC_MOCK_RESPONSE`. Test-only: never use
-  it where a real model response is required.
-- `claude` — the host's `claude` CLI, for managed sessions only (below):
-  `MOSAIC_CLAUDE_BIN` when set, else `claude` on `PATH`.
-  Not a worker adapter: the image has no `claude`, and the adapter refuses
-  without the bundle's hook and MCP files.
-
-## Managed sessions (slice 1 S6)
-
-The session runner (`packages/harness/src/runner.mjs`) runs one adapter
-call per turn on the host, as `/bin/sh <adapters/<name>/adapter.sh>` (the
-repository keeps adapters 0644; only the image sets the mode), in the
-session's workspace and its own process group. Same contract, plus:
-
-| Variable | Meaning |
-|---|---|
-| `MOSAIC_WORKSPACE` | The session's workspace; the adapter runs there. |
-| `MOSAIC_SESSION_DIR` | The session's persistent directory; later turns resume the session kept there. |
-| `MOSAIC_TOOLS` | The built-in tool limit (pi names for `pi`, Claude Code names for `claude`), comma-separated. |
-| `MOSAIC_POLICY_FILE`, `MOSAIC_TOOLS_FILE` | The bundle's gate policy and typed tools. |
-| `MOSAIC_TOOL_SOCKET` | The runner's tool socket, which the typed tools call. |
-| `MOSAIC_TURN_MARKER` | `pi`: the extension writes it at `agent_end`; a pi turn that exits 0 without it failed. |
-| `MOSAIC_EXTENSIONS` | `pi` only: extension files loaded with `-e`; a missing one exits 2. The pi module names two: the gate and typed tools extension, then the unslop check's `tools/unslop-hook/extension.ts` (row 66). |
-| `MOSAIC_UNSLOP_LOG` | The unslop check's JSONL log, `<run>/unslop.jsonl`, which the runner sets for both adapters and reads after each turn (row 66). |
-| `MOSAIC_CLAUDE_SETTINGS`, `MOSAIC_CLAUDE_MCP_CONFIG` | `claude` only: the bundle's gate hook and MCP server; required. |
-| `MOSAIC_THINKING` | The route's thinking level (row 65), which the harness module's `levels()` already checked. `pi` passes `--thinking` (`off`, `minimal`, `low`, `medium`, `high`, `xhigh`, `max`) and omits it when unset, as for `run-task.sh` workers. `claude` passes `--effort` (`low`, `medium`, `high`, `xhigh`, `max`) and requires it; unset exits 2. |
-| `MOSAIC_CLAUDE_BIN` | `claude` only: absolute path of the CLI to run, from the host's environment (the bus unit pins it). Unset or empty runs `claude` from `PATH`; a relative path or a non-executable file exits 2. |
-
-The layers each adapter relies on, and what they don't cover, are in
-`packages/harness/README.md`.
-
-### A role's skills (row 64)
-
-A managed session's skills reach both adapters as text in the bundle's
-`prompt.md`, which they pass as a file: `pi` with `--system-prompt <path>`
-and `claude` with `--system-prompt-file <path>` (row 213). Neither adapter loads a
-skill of its own: `pi` keeps `--no-skills`, since role sessions never get
-`MOSAIC_SKILLS`, and `claude` keeps `--disable-slash-commands` and
-`--restricted`. Sage chose this on 2026-10-19 from a probe, recorded in
-`agents/sage/work/queue-64/probe.md`:
-
-- Pi's `--skill` puts only a skill's name and description in the prompt,
-  and the model has to `read` the SKILL.md itself. The gate refuses every
-  file-tool path outside the workspace, so the session couldn't open its
-  own skills. Opening the gate to the bundle would change a security
-  boundary.
-- Claude Code loads a bundle-local `--plugin-dir` skill under `--restricted`
-  only with its `Skill` tool on, and then it also offers its own bundled
-  skills. Which ones it offers changes with the Claude Code version, and the
-  `Skill` tool would pass through the gate hook.
-
-The cost is prompt size: a role's skill text is in every turn, where the
-other routes load a skill when it's used. The bundle caps a role's skills at
-64 KiB together. The cap dates from when `claude` passed `prompt.md` as one
-argument, which Linux caps at 128 KiB; row 213 passes the file instead, so no
-local user reads the prompt in `/proc/<pid>/cmdline`. Pi 0.85.1 uses the path
-string itself as the prompt when the file is missing or can't be read, so the
-pi adapter exits 2 unless the path is a readable regular file.
-
-### The unslop check (row 66)
-
-Both adapters run the mechanical writing check from `tools/unslop-hook`,
-driven by one `lists.json`. `pi` loads it as its second `-e` extension.
-`claude` runs `claude-hook.js` from the bundle's `settings.json`: the Stop
-hook checks the turn's replies, and the UserPromptSubmit hook adds the
-correction notice as context on the next turn when the last reply had
-tells. Each hook command ends `; exit 0`, so neither blocks. Both log to
-`MOSAIC_UNSLOP_LOG`, and the runner adds a note for the operator to the
-next RESULT it sends (`packages/harness/README.md`, the runner). The
-harness modules mount the three files read-only at their own paths, and
-the manifest records their digests. `MOSAIC_UNSLOP_HOOK` and `UNSLOP_LISTS`
-aren't in the session environment's allowlist, so a managed session can't
-switch the check off or swap its lists.

adapters/claude/adapter.sh

@@ -1,101 +0,0 @@
-#!/bin/sh
-# Claude Code adapter: implements the Mosaic adapter contract for the host's
-# `claude` CLI, for managed sessions (slice 1 S6). Headless only. The CLI is
-# MOSAIC_CLAUDE_BIN when set (the bus unit pins it), else `claude` on PATH.
-#
-# Contract: see adapters/README.md.
-# stdout = the turn's answer only; stderr = diagnostics; exit 0 on success.
-#
-# The layers this relies on are row S0 lines 1-5 (docs/plans/2026-10-04_slice-1.md,
-# "What S6 can rely on"): the PreToolUse command hook in MOSAIC_CLAUDE_SETTINGS,
-# wrapped as `timeout -k 2 10 <gate> || exit 2`, and the tool limit (--tools).
-# --bare is never passed: it turns settings hooks off.
-set -eu
-
-need() {
-    eval "v=\${$1:-}"
-    [ -n "$v" ] || { echo "claude adapter: $1 is required" >&2; exit 2; }
-}
-need MOSAIC_SYSTEM_PROMPT_FILE
-need MOSAIC_REQUEST
-need MOSAIC_WORKSPACE
-need MOSAIC_SESSION_DIR
-need MOSAIC_MODEL
-need MOSAIC_THINKING
-need MOSAIC_CLAUDE_SETTINGS
-need MOSAIC_CLAUDE_MCP_CONFIG
-for f in "$MOSAIC_SYSTEM_PROMPT_FILE" "$MOSAIC_CLAUDE_SETTINGS" "$MOSAIC_CLAUDE_MCP_CONFIG"; do
-    [ -r "$f" ] || { echo "claude adapter: not readable: $f" >&2; exit 2; }
-done
-PROMPT_FILE=$MOSAIC_SYSTEM_PROMPT_FILE
-case "$PROMPT_FILE" in
-    /*) ;;
-    *) PROMPT_FILE="$(pwd)/$PROMPT_FILE" ;;
-esac
-[ "${MOSAIC_INTERACTIVE:-}" != "1" ] || { echo "claude adapter: interactive mode isn't supported" >&2; exit 2; }
-CLAUDE=claude
-if [ -n "${MOSAIC_CLAUDE_BIN:-}" ]; then
-    case "$MOSAIC_CLAUDE_BIN" in
-        /*) ;;
-        *) echo "claude adapter: MOSAIC_CLAUDE_BIN must be an absolute path: $MOSAIC_CLAUDE_BIN" >&2; exit 2 ;;
-    esac
-    [ -f "$MOSAIC_CLAUDE_BIN" ] && [ -x "$MOSAIC_CLAUDE_BIN" ] || { echo "claude adapter: MOSAIC_CLAUDE_BIN is not an executable file: $MOSAIC_CLAUDE_BIN" >&2; exit 2; }
-    CLAUDE=$MOSAIC_CLAUDE_BIN
-fi
-
-mkdir -p "$MOSAIC_WORKSPACE" "$MOSAIC_SESSION_DIR"
-cd "$MOSAIC_WORKSPACE"
-
-# Session: the first turn names a new session id, later turns resume it.
-# The id is kept only after a turn succeeds, so a failed first turn never
-# leaves a --resume of a session that doesn't exist.
-ID_FILE="$MOSAIC_SESSION_DIR/claude-session-id"
-if [ -s "$ID_FILE" ]; then
-    SESSION_FLAG=--resume
-    SESSION_ID=$(cat "$ID_FILE")
-else
-    SESSION_FLAG=--session-id
-    SESSION_ID=$(cat /proc/sys/kernel/random/uuid)
-fi
-
-# Flags:
-#   -p                       one-shot: print the answer and exit
-#   --output-format text     stdout is the answer only
-#   --system-prompt-file     the generated prompt replaces the default, given
-#                            as a file so no local user reads it in
-#                            /proc/<pid>/cmdline (row 213); absolute, since
-#                            the cd above changed the working directory
-#   --effort                 the instance's thinking var (row 65; 2.1.302 takes
-#                            low, medium, high, xhigh, max)
-#   --restricted             no user/project/local settings files; --settings
-#                            (the gate hook) still applies; no code-running
-#                            tool unless --tools names it; file tools confined
-#                            to the working directory; no CLAUDE.md file or
-#                            auto-memory in the prompt. The S0 lines above
-#                            don't depend on it, but it is what keeps founder
-#                            and repository memory out; a test fails without it.
-#   --tools                  the built-in tool limit (S0 line 5); empty = none
-#   --allowedTools           the same tools plus the mosaic MCP server, so
-#   --permission-mode dontAsk  nothing waits on a prompt and anything else is denied
-#   --settings               the gate hook (S0 lines 1-4)
-#   --strict-mcp-config      only the bundle's MCP server, which serves the
-#   --mcp-config             typed bus tools
-#   --disable-slash-commands no skill discovery; the role's skills are
-#                            text in the system prompt (row 64)
-ALLOWED=mcp__mosaic
-[ -z "${MOSAIC_TOOLS:-}" ] || ALLOWED="$MOSAIC_TOOLS,mcp__mosaic"
-"$CLAUDE" -p "$MOSAIC_REQUEST" \
-    --output-format text \
-    --system-prompt-file "$PROMPT_FILE" \
-    --model "$MOSAIC_MODEL" \
-    --effort "$MOSAIC_THINKING" \
-    --restricted \
-    --tools "${MOSAIC_TOOLS:-}" \
-    --allowedTools "$ALLOWED" \
-    --permission-mode dontAsk \
-    --settings "$MOSAIC_CLAUDE_SETTINGS" \
-    --strict-mcp-config \
-    --mcp-config "$MOSAIC_CLAUDE_MCP_CONFIG" \
-    --disable-slash-commands \
-    "$SESSION_FLAG" "$SESSION_ID" || exit $?
-[ "$SESSION_FLAG" = --resume ] || printf '%s\n' "$SESSION_ID" > "$ID_FILE"

adapters/mock/adapter.sh

@@ -1,18 +0,0 @@
-#!/bin/sh
-# Mock adapter: deterministic response for seam tests. NEVER use where a
-# real model response is required.
-#
-# Contract: see /opt/mosaic/adapters/README.md.
-set -eu
-
-[ -n "${MOSAIC_SYSTEM_PROMPT_FILE:-}" ] || { echo "mock adapter: MOSAIC_SYSTEM_PROMPT_FILE is required" >&2; exit 2; }
-if [ "${MOSAIC_INTERACTIVE:-}" != "1" ]; then
-    [ -n "${MOSAIC_REQUEST:-}" ] || { echo "mock adapter: MOSAIC_REQUEST is required" >&2; exit 2; }
-fi
-[ -r "$MOSAIC_SYSTEM_PROMPT_FILE" ] || { echo "mock adapter: system prompt not readable: $MOSAIC_SYSTEM_PROMPT_FILE" >&2; exit 2; }
-
-echo "mock adapter: responding verbatim from MOSAIC_MOCK_RESPONSE" >&2
-# Deterministic plumbing evidence: which MOSAIC_* variables did the
-# orchestrator actually deliver? (Auth secrets are not MOSAIC_-prefixed.)
-(env | grep '^MOSAIC_' | sort) >&2 2>/dev/null || true
-printf '%s\n' "${MOSAIC_MOCK_RESPONSE:-}"

adapters/pi/adapter.sh

@@ -1,135 +0,0 @@
-#!/bin/sh
-# Pi adapter: implements the Mosaic adapter contract for the pinned
-# @earendil-works/pi-coding-agent CLI.
-#
-# Contract: see /opt/mosaic/adapters/README.md.
-# Headless (default): stdout = response only; stderr = diagnostics; exit 0.
-# Interactive (MOSAIC_INTERACTIVE=1): full pi TUI on the attached terminal.
-set -eu
-
-[ -n "${MOSAIC_SYSTEM_PROMPT_FILE:-}" ] || { echo "pi adapter: MOSAIC_SYSTEM_PROMPT_FILE is required" >&2; exit 2; }
-# Row 213 (C2-08): pi gets the prompt as a path, so no local user reads it in /proc/<pid>/cmdline.
-# Pi 0.85.1 reads --system-prompt as a file only when the path exists and reads, and otherwise
-# uses the path string itself as the prompt, without a warning for a missing file
-# (dist/core/resource-loader.js resolvePromptInput). So the path must be a regular readable
-# file here, and is made absolute before the cd below, since pi resolves it from its own cwd.
-PROMPT_FILE=$MOSAIC_SYSTEM_PROMPT_FILE
-case "$PROMPT_FILE" in
-    /*) ;;
-    *) PROMPT_FILE="$(pwd)/$PROMPT_FILE" ;;
-esac
-[ -f "$PROMPT_FILE" ] && [ -r "$PROMPT_FILE" ] || { echo "pi adapter: system prompt not readable: $MOSAIC_SYSTEM_PROMPT_FILE" >&2; exit 2; }
-# MOSAIC_AGENT_NAME is optional in headless mode (identity section is then
-# omitted); interactive launches always set it via scripts/agent.sh.
-
-: "${PI_PROVIDER:?pi adapter: PI_PROVIDER is required}"
-: "${PI_MODEL:?pi adapter: PI_MODEL is required}"
-
-INTERACTIVE="${MOSAIC_INTERACTIVE:-}"
-if [ "$INTERACTIVE" != "1" ]; then
-    [ -n "${MOSAIC_REQUEST:-}" ] || { echo "pi adapter: MOSAIC_REQUEST is required" >&2; exit 2; }
-fi
-
-# Workspace (M5): run inside the provided workspace when present.
-if [ -n "${MOSAIC_WORKSPACE:-}" ]; then
-    mkdir -p "$MOSAIC_WORKSPACE"
-    cd "$MOSAIC_WORKSPACE"
-fi
-
-# Session (M6/M11): default ephemeral (--no-session). With a declared
-# session dir: persist there and resume the most recent session. With a
-# fork source: branch the source session file into the target dir
-# (pi --fork) - the ancestor session is never modified.
-SESSION_FLAGS="--no-session"
-if [ -n "${MOSAIC_SESSION_FORK:-}" ]; then
-    [ -n "${MOSAIC_SESSION_DIR:-}" ] || { echo "pi adapter: session fork requires MOSAIC_SESSION_DIR" >&2; exit 2; }
-    mkdir -p "$MOSAIC_SESSION_DIR"
-    SESSION_FLAGS="--fork $MOSAIC_SESSION_FORK --session-dir $MOSAIC_SESSION_DIR"
-elif [ -n "${MOSAIC_SESSION_DIR:-}" ]; then
-    mkdir -p "$MOSAIC_SESSION_DIR"
-    SESSION_FLAGS="--session-dir $MOSAIC_SESSION_DIR"
-    if [ -n "$(ls -A "$MOSAIC_SESSION_DIR" 2>/dev/null)" ]; then
-        SESSION_FLAGS="$SESSION_FLAGS -c"
-    fi
-fi
-
-# Capabilities (M5): explicit allowlist or no tools.
-TOOLS_FLAG="--no-tools"
-[ -n "${MOSAIC_TOOLS:-}" ] && TOOLS_FLAG="--tools $MOSAIC_TOOLS"
-
-# Skills (M17): explicitly provided skill dirs replace discovery. When none
-# are provided the agent runs with --no-skills (nothing ambient to find).
-# Role sessions never get MOSAIC_SKILLS (ENV_ALLOW in packages/seat/src/
-# session.mjs leaves it out), so they run with --no-skills; their skills
-# are text in the system prompt (row 64).
-SKILLS_FLAG="--no-skills"
-if [ -n "${MOSAIC_SKILLS:-}" ]; then
-    SKILLS_FLAG=""
-    OLDIFS=$IFS; IFS=','
-    for s in $MOSAIC_SKILLS; do
-        [ -d "$s" ] || { echo "pi adapter: skill dir missing: $s" >&2; exit 2; }
-        SKILLS_FLAG="$SKILLS_FLAG --skill $s"
-    done
-    IFS=$OLDIFS
-fi
-
-# Extensions (S6): explicitly provided extension files, loaded with -e
-# after --no-extensions turns discovery off. A missing file refuses here;
-# pi itself also refuses to start on a missing or broken -e (row S0 line 3).
-EXT_FLAGS=""
-if [ -n "${MOSAIC_EXTENSIONS:-}" ]; then
-    OLDIFS=$IFS; IFS=','
-    for e in $MOSAIC_EXTENSIONS; do
-        [ -f "$e" ] || { echo "pi adapter: extension missing: $e" >&2; exit 2; }
-        EXT_FLAGS="$EXT_FLAGS -e $e"
-    done
-    IFS=$OLDIFS
-fi
-
-# Thinking (row 65): a managed session's level, from its instance's thinking
-# var (pi 0.85.1: off, minimal, low, medium, high, xhigh, max). Workers from
-# run-task.sh set none and keep pi's default.
-THINKING_FLAG=""
-[ -n "${MOSAIC_THINKING:-}" ] && THINKING_FLAG="--thinking $MOSAIC_THINKING"
-
-# Mode (M13): interactive TUI or one-shot print.
-PRINT_MODE="-p"
-REQUEST_ARG=""
-if [ "$INTERACTIVE" = "1" ]; then
-    PRINT_MODE=""
-else
-    REQUEST_ARG="$MOSAIC_REQUEST"
-fi
-
-# All flags documented in the pi package README (CLI Reference):
-#   -p/--print     one-shot mode: print the response and exit (omitted in
-#                  interactive TUI mode)
-#   --system-prompt  replace the default prompt with the generated one, given
-#                  as the file's absolute path (checked above), never its text
-#   --no-*         no ambient context/skills/extensions/templates/themes
-#   EXT_FLAGS      the explicitly provided extensions only (-e)
-#   SESSION_FLAGS  ephemeral | persistent | forked (per env)
-#   TOOLS_FLAG     per capabilities
-#   THINKING_FLAG  --thinking <level> for a managed session, else nothing
-#   --offline      no startup network operations (update checks/telemetry)
-#   --no-approve   ignore project-local .pi/ files (settings, SYSTEM.md)
-#                  whatever trust is saved for the workspace
-set -- \
-    --offline \
-    --no-approve \
-    --no-extensions \
-    $EXT_FLAGS \
-    $SKILLS_FLAG \
-    --no-prompt-templates \
-    --no-themes \
-    --no-context-files \
-    $TOOLS_FLAG \
-    $SESSION_FLAGS \
-    --provider "$PI_PROVIDER" \
-    --model "$PI_MODEL" \
-    $THINKING_FLAG \
-    --system-prompt "$PROMPT_FILE"
-# One-shot mode appends -p and the request (both safely quoted);
-# interactive mode appends nothing - clean TUI.
-[ "$INTERACTIVE" = "1" ] || set -- "$@" -p "$MOSAIC_REQUEST"
-exec pi "$@"

agents/README.md

@@ -1,53 +0,0 @@
-# Mosaic Stack development team
-
-These are interactive host development agents working in the canonical
-checkout. They do not create managed fleet registrations or change role policy.
-Sage leads the project and coordinates assignments, review, and integration
-(Jason's ruling, 2026-09-26). Development sessions run in T3 for now.
-For the current bootstrap phase, direct coding, review and research use Darkwing,
-Dewey, Filbert, Rocko, Researcher and further seats Jason launches from this repository's `agents/` directory,
-not fleet seats. Keep changes in `/mnt/storage/src/mosaic-stack`; do not modify
-`~/.mosaic` launchers/provisioning or migrate/stop live fleet processes.
-
-| Agent | Responsibility | Runtime | Launch from repository root |
-| --- | --- | --- | --- |
-| Darkwing | Hands-on engineering; collaborating seat under Sage | Pi, configured Mosaic model | `agents/darkwing/launch.sh` |
-| Dewey | Frontend design, UX, accessibility, and UI implementation | Pi, configured Mosaic model | `agents/dewey/launch.sh` |
-| Rocko | General development, investigation, testing, and review | T3 session (Codex, gpt-6.1-sol, thread d99326cc, lead decision 122; Claude thread b84bb264 retired 2026-10-12); launcher Claude Code, Sonnet model | `agents/rocko/launch.sh` |
-| Filbert | General development, investigation, testing, and review | T3 session (Claude, Opus 5.5, thread 048f3c63, non-sandbox reviews only, lead decisions 130 and 134; Codex threads f712d8af and 50342f6a retired 2026-10-13, Claude thread 9cb9731e retired 2026-10-12); launcher Pi, `openai-codex/gpt-6-astra:low`, not to be used under R26 until it changes (lead decision 69) | `agents/filbert/launch.sh` |
-| Researcher | Source-grounded technical research and evidence | T3 session (Claude, Sonnet 5.5, thread 9bac0794, rows 97 and 98, lead decision 124); Haiku rows (71, 80, 84, 88, later 96) run as subagents Sage dispatches with model haiku, effort high and a worktree (lead decision 129); Haiku thread c98aaf4a retired 2026-10-13 unused; launcher Pi, configured Mosaic model | `agents/researcher/launch.sh` |
-| Launchpad | Builder, MVP T1 row 192 (Q175 SSO phase 1) | T3 session (Claude, Opus 5.5), started by Sage under mos-0974/0977 | none, T3 only |
-| Morgana | Builder, MVP T1 row 193 (human users and RBAC) | T3 session (Claude, Opus 5.5), started by Sage under mos-0974/0977 | none, T3 only |
-| Honker | Builder, MVP T1 row 194 (agent lifecycle) | T3 session (Claude, Opus 5.5), started by Sage under mos-0974/0977 | none, T3 only |
-| Gander | Builder, the human boundary across uids (P19 333, #1694) | T3 session (Codex, gpt-6.1-sol), started by Sage under mos-0977 | none, T3 only |
-| Herb | Builder, a second gate lane (#1695) | T3 session (Codex, gpt-6.1-sol), started by Sage under mos-0977 | none, T3 only |
-| Bushroot | Builder, retire the control board writer on 7331 (#1696) | T3 session (Claude, claude-haiku-5-5), started by Sage under mos-1008 | none, T3 only |
-| Megavolt | Builder, MVP T1 item 6 (agent credentials and git identity, #1697) | T3 session (Claude, Opus 5.5), started by Sage under mos-0974/0977 | none, T3 only |
-| Quackerjack | Builder, federation R5 and R2 (#414) | T3 session (Claude, claude-haiku-5-5), started by Sage under mos-1023 | none, T3 only |
-| Gizmoduck | Federation design (#414) and builder of FED R1 | T3 session (Claude, Opus 5.5, thread 9312a896) | none, T3 only |
-| Webby | Reviewer: Case 2 designs, rows 192 and 194 | T3 session (Claude, Opus 5.5, thread 03086006), started by Mos (mos-0981) | none, T3 only |
-| Duckworth | Reviewer: federation designs, row 193 | T3 session (Claude, Opus 5.5, thread ffceafec), started by Mos (mos-0977) | none, T3 only |
-| Sage | Project lead: coordination, review, integration; earlier DYOR strategy records retained | T3 session (Claude Code); Pi launcher `zai/glm-5.3:high` retained | `agents/sage/launch.sh` |
-
-Each script supports `--check` and `--fresh`. Normal launches resume the agent's
-own conversation; a first launch starts one. See each agent's README for
-context inputs, authentication, and recovery details. Launch scripts can also
-be invoked by absolute path from any directory. No assignment or model request
-is submitted by the launcher itself.
-
-The shared Pi helper supports `--provider NAME`, `--model ID`, and
-`--thinking LEVEL` as per-launch overrides of the validated system defaults.
-Filbert's wrapper appends the required provider, model and thinking flags so
-its launch configuration remains fixed, including on resume. Use Filbert's
-wrapper to select that configuration; a direct shared-helper invocation uses
-its own supplied flags or the system defaults.
-
-All agents follow repository governance and current user direction. Team
-leadership does not add deployment or push authority. Coordinate overlapping
-work with Sage and preserve other sessions' changes.
-
-Before 2026-09-26 Sage worked only on DYOR strategy and sat outside the
-development queue. Jason then made Sage the project lead and Darkwing a
-collaborating seat. A separate fleet Sage seat under `~/.mosaic` is being
-decommissioned; it does not speak for this seat. Whether the DYOR strategy work
-continues is Jason's call. Joe retains DYOR engineering.

agents/bushroot/README.md

@@ -1,8 +0,0 @@
-# Bushroot
-
-Builder seat, Claude Haiku 5.5, started by Sage under mos-1008 item 2 (a low-level model under Q177) for the row
-that retires the control board writer on 7331 (#1696).
-
-No launcher in this directory. The seat runs as a T3 thread and works in its own worktree. It has no
-credentials of its own; until the credentials row settles per-agent identity, it pushes only its own `bushroot/*`
-branches through the repository credential helper (mos-0981), never `refactor`, `main` or another seat's branch.

agents/darkwing/CONTEXT.md

@@ -1,38 +0,0 @@
-
-===== DARKWING NATIVE DEVELOPMENT CONTEXT =====
-
-Your identity is Darkwing. This launch runs Pi directly on the host, in the
-Mosaic Stack development repository. The injected SOUL defines your persona;
-CONSTITUTION and STANDARDS supply governance, USER supplies user context,
-and AGENTS.md supplies repository instructions.
-
-You have host read, bash, edit, write, grep, find, and ls tools. This is a
-development TUI with the operator's OS access, not a sandbox or a registered
-managed fleet seat. Use repository scripts for Mosaic operations and inspect
-their effects before running them. Container paths in skills describe worker
-deployments, not your current workspace. A tool's presence is not authority
-to change unrelated files, other agents' work, or the live fleet.
-
-For an assigned improvement, inspect the implementation, reproduce the issue,
-make the smallest useful change, verify it, and continue through the authorized
-outcome. Run scripts/mosaic queue next darkwing for ownership, gates and your next piece;
-a new user assignment does not silently resume unrelated queued work.
-
-The local /goal extension is loaded and owns any operator-set goal lifecycle.
-Use ms-proactive-agent for work selection and ms-goal for recovery guidance;
-do not create a competing goal loop. Follow goal_report's actual schema and
-reporting instructions. Its text format is Just Completed / Next Step /
-Blocked, with '* none' for empty sections. No external reporting skill is
-needed to discover that format. Native development packaging supersedes
-older skill statements that this extension is unavailable.
-
-For relocation recovery, read agents/darkwing/work/RESTART.md after the root
-AGENTS.md and docs/plans/CURRENT.md. It records verified checkpoints and limits,
-not a new assignment. The canonical checkout is /mnt/storage/src/mosaic-stack;
-v1/ is archived legacy source. Reconcile newer owner direction before acting.
-
-Conversation history persists across launcher restarts. Goals belong to a
-single process incarnation; recover the assignment from verified records and
-the operator's direction after a restart. No goal is started by this launcher.
-Context is captured anew at launch; source edits do not update this process's
-injected snapshot. Relaunch to load approved context changes.

agents/darkwing/README.md

@@ -1,64 +0,0 @@
-# Darkwing development TUI
-
-From any terminal, run:
-
-```sh
-/home/jwoltje/src/mosaic-stack-dev-test/agents/darkwing/launch.sh
-```
-
-The agent launcher is a thin shim to `scripts/agent.sh --host-dev darkwing`,
-forwarding all arguments unchanged. `scripts/agent.sh` is the common entry
-point; `scripts/agent-host-dev.sh` implements its native development mode.
-The host launcher opens the repository as Darkwing's workspace.
-It uses the repository-pinned Pi, the configured Mosaic provider/model, and
-native Pi authentication (normal `~/.pi/agent`, or `PI_CODING_AGENT_DIR` if
-explicitly set). It never copies credentials. Install dependencies with
-`npm ci --ignore-scripts --no-audit --no-fund` if needed.
-
-`--check` validates configuration and required inputs without opening Pi or
-calling a model. `--fresh` starts a new conversation without deleting earlier
-ones. Normal launches continue the latest conversation under
-`.pi/state/darkwing/sessions/`; the first launch creates one. A launcher lock
-rejects simultaneous launches through this script. It does not exclude Pi
-processes started another way. Damaged JSONL history refuses automatic resume;
-`--fresh` is an explicit escape hatch that preserves the damaged evidence.
-
-The current files are combined into a private launch snapshot under
-`.pi/state/darkwing/launches/`:
-
-- `contracts/CONSTITUTION.md` and `contracts/STANDARDS.md`
-- `agents/darkwing/SOUL.md`
-- `<configured dataRoot>/user/USER.md`, the deployment's live user profile
-- the repository's `AGENTS.md` and Darkwing's `CONTEXT.md`
-
-Use `--soul FILE`, `--constitution FILE`, or `--user FILE` to select alternate
-inputs, including a future `contracts/USER.md`. Relative paths resolve from
-the repository root. Missing or empty inputs refuse launch. Snapshots can
-contain personal context and remain local, with private file permissions.
-Context edits take effect on relaunch, including when resuming a conversation.
-
-The launcher enables coding/search tools, `goal_report`, ten explicit local
-skills, and the canonical goal extension through `scripts/sync-dev-extensions.sh`.
-Ambient context, skills, extensions, templates, and themes are disabled.
-The normal Pi coding prompt is retained with the Mosaic context appended.
-Enter `/goal <assignment and acceptance criteria>` to start continuing work;
-`/goal stop`, `/goal resume`, and `/goal` pause, resume, and inspect it. A new
-process does not automatically adopt a previous process's goal.
-
-This TUI has the operator's host access, including repository edits and host
-commands. Its tool list is not OS isolation. It creates no managed role or
-fleet registration. Worker dispatch still uses the governed Mosaic task runner.
-The user supplies the assignment; launch alone does not start self-modification.
-
-## Deployment findings
-
-The existing `scripts/agent.sh` launches a Docker container, defaults to the
-`agent-<name>` session directory, and asks Pi to continue when that directory
-is nonempty. Its default workspace is `<dataRoot>/workspaces/<name>`, not this
-checkout. `src/load-contracts.sh` loads image-baked governance, an optional
-seat SOUL override, live user Markdown, and mission context into a shared
-prompt path. A seat override requires `agent.json`; a standalone SOUL is not
-discovered. `adapters/pi/adapter.sh` disables extensions. The temporary host
-launcher follows the existing native development path to provide repository
-access and `/goal`, and keeps its conversations separate from container and
-live fleet sessions. It does not invoke release alignment on startup.

agents/darkwing/SOUL.md

@@ -1,30 +0,0 @@
-# SOUL — Darkwing
-
-You are Darkwing, Mosaic Stack's hands-on engineering collaborator, working
-with Sage as project lead. Your job is to help Jason make the system
-dependable by using it, finding where it falls short, and carrying authorized
-improvements through verification.
-
-Be curious, direct, and resourceful. Have a technical opinion and explain
-the evidence behind it. Investigate before guessing. Distinguish a design
-claim, a passing test, and behavior you have observed in the running system.
-
-Use Mosaic's own tools and workflows where they fit. Turn a failure into a
-reproducible case, make a focused correction, and test the behavior again.
-Let each verified improvement inform the next one within the assignment.
-Keep the human informed when the result, scope, or next decision changes.
-
-Own the outcome while respecting other agents' work. Preserve their changes
-and records, give delegated work clear boundaries, and seek independent
-review where required. Self-improvement never grants new authority: changing
-your instructions, permissions, or a live deployment follows the same review
-and authorization rules as any other system change.
-
-Sage leads the development team (Jason's ruling, 2026-09-26) and translates
-Jason's priorities into scoped work, coordinates ownership and dependencies,
-reviews results, and verifies integration. You are a collaborating seat.
-Dewey owns frontend design and UX. Rocko and Filbert support general project needs,
-including implementation, investigation, testing, and review. Reconcile
-concurrent edits with Sage before integration. Keep Jason informed of
-outcomes and decisions that require his input. Your role does not expand the
-project's existing authorization or release rules.

agents/darkwing/launch.sh

@@ -1,10 +0,0 @@
-#!/usr/bin/env bash
-# Darkwing's native development mode through the Mosaic agent entry point.
-set -euo pipefail
-REPO="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
-# Register this seat with the control board (packages/seat) unless already
-# registered by `mosaic launch` or only running the checks.
-if [ -z "${MOSAIC_LAUNCH_REGISTERED:-}" ] && ! printf '%s\n' "$@" | grep -qx -- '--check'; then
-    exec "$REPO/scripts/mosaic" launch --repo "$REPO" --harness pi darkwing -- "$@"
-fi
-exec "$REPO/scripts/agent.sh" --host-dev darkwing "$@"

agents/darkwing/validate-sessions.mjs

@@ -1,21 +0,0 @@
-// Refuse damaged history before Pi's --continue can silently skip it.
-import { readFileSync, lstatSync } from 'node:fs';
-
-try {
-  for (const file of process.argv.slice(2)) {
-    if (!lstatSync(file).isFile()) throw new Error(`not a regular session file: ${file}`);
-    const lines = readFileSync(file, 'utf8').trim().split('\n');
-    const entries = lines.map((line) => JSON.parse(line));
-    const header = entries[0];
-    if (header?.type !== 'session' || typeof header.id !== 'string' || !header.id ||
-        typeof header.version !== 'number' || typeof header.cwd !== 'string' ||
-        !Number.isFinite(Date.parse(header.timestamp)) ||
-        entries.slice(1).some((entry) => !entry || typeof entry.type !== 'string')) {
-      throw new Error(`invalid session structure: ${file}`);
-    }
-    if (header.cwd !== process.cwd()) throw new Error(`session belongs to another workspace: ${file}`);
-  }
-} catch (error) {
-  console.error(`darkwing: cannot safely resume: ${error.message}; inspect history or explicitly use --fresh`);
-  process.exit(1);
-}

agents/darkwing/work/RESTART.md

@@ -1,108 +0,0 @@
-# Darkwing — relocation handoff
-
-Recorded 2026-09-07 17:43 UTC. Jason intends to relaunch with
-`/mnt/storage/src/mosaic-stack/agents/darkwing/launch.sh`.
-This is a recovery note, not a new assignment or automatic goal resumption.
-
-## Read first
-
-1. Root `AGENTS.md` and `docs/plans/CURRENT.md`.
-2. This note, then `git status --short` and `git log --oneline -5`.
-3. Reconcile current owner direction and any newer declared artifacts before acting.
-
-## Repository conversion is completed locally
-
-Jason explicitly ordered the conversion and confirmed no work was active.
-- Canonical checkout: `/mnt/storage/src/mosaic-stack`.
-- Origin: `https://git.mosaicstack.dev/mosaicstack/stack`.
-- Branch: `refactor`.
-- Conversion commit: `127a54fdff1fe6ae56c3197edddf957481465db4`.
-- New foundation is at root. `v1/` is legacy archival source, NOT current code.
-- Old `/home/jwoltje/src/mosaic-stack-dev-test` is a compatibility symlink to this
-  same checkout. Do not recreate a second working copy there.
-- Both histories retained: merge parents v2 `9a5fbdbda74b16adf488fe28138b2ba69ea5e669`
-  and v1 `5d2770002612a09ae0cadc129b4ea30619133e8a`.
-- Exact 3,507-file v1 tracked tree imported; v1 refs under `refs/archive/v1/`.
-- Original v2 refs retained; `stack-v2-archive` remote has a disabled push URL.
-- Only legacy tracked tree and four conversion docs committed. All earlier
-  uncommitted/untracked/ignored work preserved. Index was verified clean.
-- Issue https://git.mosaicstack.dev/mosaicstack/stack/issues/1495 closed explicitly
-  for local conversion. No push, PR/trunk merge or live-service change occurred.
-
-Record: `docs/plans/2026-09-07_repository-consolidation-completed.md`.
-Receipts: `docs/plans/reviews/2026-09-07_repository-conversion-verification.json`
-and `2026-09-07_repository-conversion-postcommit-verification.json`.
-Verified rollback copies, NOT development roots:
-- `/mnt/storage/src/.mosaic-stack-conversion-20260907T172430Z/`
-- `/home/jwoltje/src/.mosaic-stack-dev-test.pre-conversion-20260907T172430Z`
-Do not delete them, launch from them or restore over newer work.
-
-## Current unfinished foundation gate
-
-Jason's A9 acceptance of the first offline synthetic scope/permission inspector
-is pending. Code is independently approved by Filbert; no blocking code finding
-remains at the reviewed r6 candidate. Owner acceptance is not inferred from tests.
-
-- Manifest: `docs/plans/reviews/2026-09-07_foundation-inspector-rocko-build-manifest-r6.json`
-  SHA-256 `a4a4493000aff5905337a643886ca36e7c5377d52deed77b8aeab7174ca73dcf`.
-- Report: `docs/plans/reviews/2026-09-07_foundation-inspector-rocko-build-r6.md`
-  SHA-256 `ee0e83efd7c71eddecf5e26f939e9a34ba85b184cfcd1cffac9ff9e56ea13c37`.
-- APPROVED verdict: `docs/plans/reviews/2026-09-07_foundation-inspector-code-verdict-r6.md`
-  SHA-256 `ab9dd5e5c3cad5c9263e873ff82cac444da2d36040e907e4798b208fa1c08b13`.
-- Guide: `docs/plans/reviews/2026-09-07_foundation-inspector-demo.md`.
-
-All 382 approved inspector files and pinned inputs survived conversion unchanged.
-Actual offline checks: Node 80/0, selftests 43/0, oracle 1,568 records / zero
-schema disagreements, foundation checker PASS, config/auth/conductor 24/15/17.
-Postcommit conductor 17/0 and four CLI demos passed: allowed read, allowed change
-PREVIEW (no mutation), missing-registration refusal, unresolved reassignment with
-original selection retained. Demo inputs are separate synthetic scenarios.
-
-`test-task.sh` and `test-release.sh` remain NOT RUN / DEFERRED under Jason's bounded
-offline-demo ruling. No deployment/native/live/provider/security certification.
-Reviewer qualifications: ordering equality means structural equality, not byte
-identity; auxiliary native-parser warm-run anomalies remain separate unresolved
-observations, not a passing universal parser-equivalence claim. Preserve all earlier
-NOT APPROVED reviews and the historical correction that r3 ran unauthorized live
-branches; later deferral did not retroactively authorize them.
-
-## Ownership and limits
-
-- Rocko authored inspector code; Filbert independently reviewed; Darkwing coordinates
-  and verifies. Keep the approved candidate frozen unless a new fix is authorized.
-- No automatic permission to push, merge to next/main, deploy, change live config,
-  grant permissions, access credentials, investigate ~/.mosaic, or start new runtime
-  work. Local conversion authority is not authority for those activities.
-- Preserve unrelated pending work. In particular `scripts/agent.sh`, `docs/TOOLS.md`,
-  host launcher/context files and other untracked concepts/skills belong to existing
-  work. Do not blanket-stage/reset/clean. Root logs and CURRENT remain uncommitted.
-- Foundation #53 in the old stack-v2 project remains a separate open issue; do not
-  silently close or renumber it. Accepted historical SHA/path citations remain valid.
-- Rocko's Archify C1 remains HELD for owner T2/T3 decisions. No lane reassignment.
-- Future durability/workflow/evidence/federation/onboarding topics are notes, not
-  authorization to expand the inspector.
-
-## Communications
-
-Use only `tools/tmux/agent-send.sh`; sender `dragon-lin:darkwing`.
-Rocko: `-L mosaic-fleet -s '=rocko'`; Filbert/Dewey:
-`-L default -s '=filbert'` / `'=dewey'`.
-Conversion notice delivered to Rocko. Filbert/Dewey sends were unconfirmed
-(input boxes not locatable); no retries, no acknowledgement claimed. Check declared
-artifact paths as well as direct messages; completed reviews have existed without
-transported replies. Do not inspect private panes or blindly resend.
-
-## Relaunch and goal recovery
-
-The project launcher continues its own latest `.pi/state/darkwing/sessions/`
-conversation by default. Do NOT assume this pre-launch conversation is already in
-that store or that the next launch resumes this exact conversation. This handoff
-is the durable bridge. No session-tree migration or launch was performed here.
-
-The goal extension owns lifecycle. The earlier extension goal had been paused;
-no restart automatically resumes it. Reconcile the actual new process state and
-Jason's direction rather than reporting progress against a guessed old goal or
-creating a second goal loop. Launch alone grants no new assignment.
-
-This handoff and its CONTEXT pointer are documentation-only. Launcher scripts,
-private sessions, credentials and runtime configuration were not modified.

agents/darkwing/work/board-attention/r1.json

@@ -1,17 +0,0 @@
-{
-  "issue": 1503,
-  "candidate": "/tmp/board-attention-r1-q_924ksh",
-  "files": [
-    "AGENTS.md",
-    "packages/control-board/src/scan.mjs",
-    "packages/control-board/README.md",
-    "packages/control-board/tests/scan.test.mjs",
-    "packages/control-board/tests/serve.test.mjs",
-    "packages/control-board/tests/attention.test.mjs",
-    "packages/control-board/tests/attention-flow.test.mjs",
-    "packages/webui/tests/fixture.mjs",
-    "docs/plans/2026-09-13_board-attention-status.md"
-  ],
-  "manifestSha256": "e40b58ecb6844d407ba776dcde8f19ad21c0b78076ca1f9b3d96b0bb1c852405",
-  "testsPassed": 144
-}

agents/darkwing/work/board-attention/restart-verification.json

@@ -1,14 +0,0 @@
-{
-  "at": "2026-09-14T00:19:32.409806+00:00",
-  "backendPid": 3204655,
-  "health": "ok",
-  "researcher": {
-    "agent": "researcher",
-    "project": "mosaic-stack",
-    "alive": true,
-    "state": "idle",
-    "waitingOnYou": false,
-    "lastActivity": "2026-09-13T21:19:57.102Z"
-  },
-  "fiveAgentProcessesUnchanged": true
-}

agents/darkwing/work/board-attention/restart.json

@@ -1,46 +0,0 @@
-{
-  "at": "2026-09-14T00:19:11.579013+00:00",
-  "ownerAuthorized": true,
-  "oldPid": 1265952,
-  "newPid": 3204655,
-  "command": [
-    "/usr/bin/node",
-    "packages/control-board/src/cli.mjs",
-    "serve"
-  ],
-  "cwd": "/mnt/storage/src/mosaic-stack",
-  "log": "/tmp/board-attention-backend-ag9xvks2.log",
-  "agentProcessesBefore": {
-    "default/darkwing": [
-      [
-        "2733924",
-        "12863634"
-      ]
-    ],
-    "default/dewey": [
-      [
-        "934346",
-        "466065"
-      ]
-    ],
-    "default/filbert": [
-      [
-        "72183",
-        "100870"
-      ]
-    ],
-    "default/researcher": [
-      [
-        "173699",
-        "66404285"
-      ]
-    ],
-    "mosaic-fleet/rocko": [
-      [
-        "90599",
-        "128275"
-      ]
-    ]
-  },
-  "gracefulExitObserved": true
-}

agents/darkwing/work/chat-02-routes-review-2026-09-26.md

@@ -1,54 +0,0 @@
-# CHAT-02 board routes: Darkwing's review (#1507)
-
-Reviewer: Darkwing, 2026-09-26, per Sage's D3. Requested by Dewey. Scope: the
-two read-only routes only. Filbert reviews `packages/conversation` in full.
-
-Candidate, base 34777c56, uncommitted. I verified both hashes:
-- `packages/control-board/src/serve.mjs` afc95bdb…c540d
-- `packages/control-board/tests/serve.test.mjs` e60aa14b…ecbc
-
-**Verdict: approve**, with one commit condition and two nonblocking notes.
-
-## What I checked
-
-- Order. `foreignRequest` runs first on every request, then the POST routes,
-  then the GET/HEAD check (405 otherwise), then these routes. A POST to
-  either path is 405, and a foreign Host or Origin is 403 before any read.
-- Query validation. `/api/conversations` refuses any parameter.
-  `/api/conversation` accepts only `id`, `branch` and `cursor`, one value
-  each, each matching `QUERY_VALUE`. That is the same pattern as
-  `parts.mjs` `ID`, so every id the reader issues (`safeId`, `root`, `c-`
-  cursors, `pi-` conversations) passes. No path comes from the request.
-- Responses. JSON with `no-store` and `nosniff`, no CORS headers. A thrown
-  error gives a fixed 500 body and logs to stderr only.
-- Refusal bodies. Every `Refusal` message in `packages/conversation/src` is
-  a fixed string. The one interpolated message (`denied`, safe-fs.mjs:34)
-  interpolates only "session root" or "session file". No path or content
-  reaches the client through `error`.
-- Status map. It covers every code the route can reach. `unknown-actor` and
-  `unsupported-purpose` are absent, and the route can't produce them because
-  it always passes the default actor and purpose.
-- Tests: `serve.test.mjs` plus `packages/conversation/tests/`, 61/61 on the
-  pinned files.
-
-## Commit condition
-
-`serve.mjs` imports `../../conversation/src/reader.mjs` at module load, and
-`packages/conversation/` is untracked. Committing the routes without that
-package breaks the board's start, not only these routes. The package must
-land in the same commit or an earlier one, after Filbert's review.
-
-## Notes (nonblocking)
-
-1. **A cursor needs its branch.** The header comment says `branch` and
-   `cursor` are optional. But `next()` compares `branch !== record.branch`,
-   and every cursor record carries a string branch (`safeId` or `root`). So
-   `?id=X&cursor=C` without `branch` is always 409 `cursor-foreign`, with
-   `reconcile: true`. That is safe, but a client that follows `nextCursor`
-   alone gets a refusal that reads like a stale view. The test passes the
-   page's branch, so it doesn't show this. Either say in the comment that a
-   cursor call must repeat `page.branch`, or answer 400 "cursor requires
-   branch". I'd take the comment now and let CHAT-03's client decide.
-2. **New codes fall to 422.** A code the reader adds later maps to 422
-   without a test failing. A test that runs the reader's refusal codes
-   through `REFUSAL_STATUS` would catch that. That's optional.

agents/darkwing/work/chat-02-routes-review-r2-2026-09-26.md

@@ -1,69 +0,0 @@
-# CHAT-02 board routes, revision 2: Darkwing's review (#1507)
-
-Reviewer: Darkwing, 2026-09-26, at Sage's request. Scope: the route delta
-since my R1 approval (`chat-02-routes-review-2026-09-26.md`, 07b10ad1). Filbert
-reviewed the backend (packet `agents/dewey/work/chat-02/BACKEND.md`, 0cf177b1).
-
-Candidate, base 34777c56, uncommitted. I verified both hashes:
-- `packages/control-board/src/serve.mjs` d62720dc…a2f3
-- `packages/control-board/tests/serve.test.mjs` d38aa2b2…3f4a
-
-**Verdict: approve.** The R1 commit condition still holds, and I have two
-new nonblocking notes.
-
-## What I checked
-
-I kept no copy of the R1 files, so I read the whole route change against the
-base (`git diff 34777c56 -- packages/control-board`) instead of only the
-delta. It covers every item the packet's §0 lists and nothing else in the
-route path.
-
-- **Cursor needs its branch.** This was my R1 note 1. `conversationQuery` now
-  answers 400 "a cursor call repeats the page's branch" when `cursor` comes
-  without `branch`. The check runs after the per-key validation, so a
-  malformed value still gets its own 400 first. The header comment says the
-  same. A test covers it, and removing the line fails it.
-- **Status map.** My R1 note 2. `REFUSAL_STATUS` is exported and now has 16
-  entries. I listed every `new Refusal("<code>"` in
-  `packages/conversation/src` myself and got 15 codes plus
-  `unsupported-harness`, which reader.mjs:352 raises by value. That matches the
-  map exactly. `parts.mjs` raises none. `unavailable`, which safe-fs.mjs:58
-  raises when a session root doesn't exist, is 404. That fits the rest of the
-  map, where not-found is 404. `unknown-actor` 403 and `unsupported-purpose` 422
-  are explicit now.
-- **Order and guards** are unchanged from R1. The foreign Host or Origin check
-  comes first, then the POST routes, then 405, then these routes. No path comes
-  from the request, and responses carry `no-store` and `nosniff` with no CORS
-  headers.
-- **Tests.** `serve.test.mjs` plus `packages/conversation/tests/` pass
-  67/67 on the pinned files.
-- **Mutations** on a scratch clone of HEAD with the conversation package and
-  the two pinned files:
-
-  | Mutation | Result |
-  |---|---|
-  | cursor-without-branch check removed | 1 fails |
-  | `unknown-actor` entry dropped | 1 fails (the scan test) |
-  | `nosniff` removed | 1 fails |
-  | repeated-parameter check removed | 1 fails |
-  | catalogue parameter check removed | 1 fails |
-  | `unavailable` changed from 404 to 422 | nothing fails |
-
-  The last row is note 1 below.
-
-## Commit condition (unchanged)
-
-`serve.mjs` imports `../../conversation/src/reader.mjs` at module load, and
-`packages/conversation/` is still untracked. The package must land in the
-same commit as the routes or an earlier one. Otherwise the board fails to
-start.
-
-## Notes (nonblocking)
-
-1. **The scan test checks keys, not values.** It proves every code has an
-   entry. No test proves `unavailable` is 404. If someone edits that value, or
-   any status no route test exercises, nothing fails. A table test that
-   asserts the whole `REFUSAL_STATUS` object would pin them. That's optional.
-2. **The scan reads a fixed list of three files.** If a refusal is added to
-   `parts.mjs` or a new file, the scan won't see it, and that code falls to 422.
-   Reading every `.mjs` in `packages/conversation/src` would close the gap.

agents/darkwing/work/chat-03-I1-review-r1-2026-10-04.md

@@ -1,160 +0,0 @@
-# Queue row 5, CHAT-03 increment 1, round 1 review (#1508)
-
-Darkwing, 2026-10-04. Request: #1508 comment 26671. My part is the binding
-and the extension-load refusal (lead decisions 31 and 36). Brief:
-`agents/dewey/work/chat-03/BRIEF.md` at 1ef15ac0. Candidate:
-`agents/dewey/work/chat-03/I1-manifest.sha256`, digest
-`1404341eaeaf7d1e684c9f27e76718061ed08c25a52da5f190425feb1274ba69`,
-27 files, uncommitted.
-
-Verdict: changes requested. Two blocking findings, B1 and B2.
-
-## Checks
-
-- The manifest hashes to 1404341e. All 27 files match it in the canonical
-  working tree.
-- Export: `git archive` of 1c724958 plus the 27 candidate files in
-  `/tmp/r5-exp`, manifest OK there too. `node --test
-  packages/conversation/tests/` passes 141/141, 0 skipped.
-  `node --test packages/control-board/tests/` passes 124/124 (the board
-  reads `ControlRefusal` codes). The CHAT-00, CHAT-01 and CHAT-01c checks
-  still pass.
-
-## B1 (blocking). The seal is a deny-list over an argv the caller builds
-
-`checkSeal` (pi-pin.mjs 59–66) refuses `-e`/`--extension`, a missing seal
-flag, or a first pair other than `--mode rpc`. Everything else in
-`engine.extraArgs` passes, and `buildPiArgs` appends extraArgs after the
-controller's own `--session`. Pi's parser keeps the last `--mode` and the
-last `--session`. The controller is a public export (`./controller` in
-package.json), so this is reachable without touching the source.
-
-(a) `extraArgs: ["--session", <file outside the fixture root>]`. The guard
-refuses that same path when it is passed as `sessionFile`, but it never
-sees extraArgs. With the real pinned Pi under a scratch HOME and agent dir
-(`/tmp/r5/seal-escape.mjs session`):
-
-```
-guard on sessionFile: live-session-refused
-constructed with extraArgs ["--session",".../outside/proj/.pi/state/other/sessions/s1.jsonl"]
-  -> piArgs ["--mode","rpc","--no-extensions","--no-prompt-templates","--no-themes",
-             "--session",".../fx/.../fixture-seat/sessions/s1.jsonl","--session",".../outside/..."]
-start -> {"launched":true,"classified":{"state":"free"}} | binding uncertain closed
-uncertain evidence: loaded-session, "the engine loaded another session file"
-outside session changed: true | appended: {"type":"thinking_level_change","id":"bef67aef",
-  "parentId":"b2c3d4e5",...,"thinkingLevel":"off"}
-```
-
-K8 notices afterwards, but Pi has already written to a session the guard
-exists to protect. That breaks §2 "no writes to sessions" and the
-fixture-only rule in code.
-
-(b) `extraArgs: ["--mode", "json"]` (or `text`). checkSeal passes because
-it looks only at args[0] and args[1]. Pi starts in print mode, reads stdin
-to EOF and treats it as the prompt. The K8 `get_state` line goes into that
-reader, so the run ends `uncertain` on `get_state: timeout`. In an earlier
-run where I closed stdin, Pi sent the `get_state` JSON line to the model
-as a prompt and stopped only at "No API key found". The default
-`engine.env` is `process.env`, so with real auth present that becomes a
-paid model call outside RPC. PgroupLauncher spawns detached, so the engine
-outlives the controller unless someone kills the group.
-
-Other extraArgs that pass checkSeal today: `--no-session`, `--fork`,
-`--export <file>` (writes a file), `--prompt-template`, `--approve`.
-`engine.command` and `engine.preArgs` can also run any script, or node
-with `--import`. `checkEnginePin` validates only the pinRoot lock files, so
-the pin says nothing about what actually ran.
-
-Suggested fix:
-- Allow-list extraArgs. The only non-extension use in the suites is
-  `["--model", "other"]` (claim.test 548, W9), so `--model`, `--provider`
-  and `--thinking` with one value each would cover it.
-- Refuse any second `--mode`, `--session` or seal flag, and any session
-  or output flag (`--print`, `--no-session`, `--session-dir`,
-  `--session-id`, `--fork`, `--export`, `--continue`,
-  `--resume`).
-- Say in the README that a non-default `command` or `preArgs` is a test
-  hook, and that the pin and seal checks don't bind under it. Or refuse it
-  outside tests.
-- N24 cases for `--session`, `--mode json` and `--no-session` in
-  extraArgs.
-
-## B2 (blocking). The claim's session key is the conversation ID
-
-controller.mjs 187:
-`this.sessionK = sessionKey({ harness: "pi", conversation: this.conversation })`.
-`conversation` is `"pi-" + sha256(projectRoot, seat, name)` (reader.mjs
-72). The brief (line 382–383) keys the claim on "the native session
-identity (the Pi header ID or the Claude session UUID)", and the CHAT-01
-README says at most one non-stopped binding may hold a conversation or
-native-session identity. Two paths to one session file give two
-conversation IDs, so the session key never collides.
-
-Repro, `/tmp/r5/hardlink.mjs`: one session hard-linked into
-`.pi/state/fixture-seat/sessions/s1.jsonl` and
-`.pi/state/seat-b/sessions/s1.jsonl`, two controllers via the test harness
-with the fake engine.
-
-```
-same inode: true
-A conversation pi-b5901a69... | B conversation pi-f21f1a75...
-A start: {"launched":true,...} state active
-B start: {"launched":true,...} state active
-A nativeSession 0f5e1c2a-1111-4222-8333-944455556666 | B nativeSession 0f5e1c2a-1111-4222-8333-944455556666
-```
-
-Two active bindings, two engines, one session. A plain copy is allowed
-too; whether a copy should count as the same session is a call for the
-brief, but the hard link plainly should. W4 (claim.test 172) builds its
-keys by hand on the store, so it never exercises the controller's
-mapping.
-
-Fix: build the session key from the header `id` read in `start()` (the
-`nativeSession` already in hand). Add a controller-level W4 case for the
-hard link, and one for a copy with whatever the brief decides.
-
-## n1 (non-blocking). The startup-append note is narrower than Pi's rule
-
-README 426–431 and BUILD-I1.md 196 say the append bites only sessions
-without a `thinking_level_change` entry, and that Pi-created sessions
-carry one. Pi's rule is sdk.js 82:
-`hasExistingSession = existingSession.messages.length > 0`. A session
-with a thinking entry but no messages takes the new-session branch and
-appends `thinking_level_change` at every start, plus `model_change` when a
-model is set. I checked this in plain sealed RPC mode: a header plus a
-thinking entry gained `{"type":"thinking_level_change","id":"e8c74875",
-"parentId":"f0e1d2c3",...}`. A Pi-created session that was opened and
-never prompted is in this group, so "written by something else" is wrong.
-Name message-less sessions too, and let the later pre-spawn check cover
-both.
-
-## n2 (minor). The guard follows $HOME
-
-`LiveSessionGuard` protects `~/.pi`, `~/.claude` and `~/.mosaic-dev` via
-`os.homedir()`. With a scratch HOME, the real directories are protected
-only by the fixture-root containment, or by passing `homes`. That
-containment holds today, so I'm noting it, not blocking on it.
-
-## What holds in the binding
-
-- H1 to H3. `#dispatch` holds `this.lock` across `#recheck` and the
-  write. Takeover, release and acquire run `#evaluate` under the same
-  lock, so a prompt can't land between the generation bump and the write.
-- The generation check comes first in `#evaluateOp` and again in
-  `#recheck`.
-- Takeover refuses while fenced (K9) and for the current controller
-  (`already-controller`, H4).
-- Disconnect never moves control (H11).
-- Confirmations are single-use: `#checkConfirmation` marks them
-  `consumed`.
-- Interrupt sets its fence before it takes the lock, so a queued prompt
-  sees the fence.
-- K8 catches a wrong session or leaf after launch, as in B1(a). B1 is that
-  the write happens before K8 can run.
-- The pin check reads both lock files and refuses on either version or
-  integrity mismatch.
-
-Scratch scripts and outputs are in `/tmp/r5/`: `seal-escape.mjs`,
-`hardlink.mjs`, `seal-session.txt`, `seal-json.txt`, `hardlink.txt`,
-`conv-suite.txt`, `board-suite.txt`. All scratch engines were killed by
-their process group.

agents/darkwing/work/chat-03-I1-review-r2-2026-10-04.md

@@ -1,119 +0,0 @@
-# Queue row 5, CHAT-03 increment 1, round 2 review (#1507)
-
-Darkwing, 2026-10-04. Request: #1507 comment 26689 (Dewey). My part is
-the binding and the extension-load refusal (lead decisions 31 and 36), and
-my round 1 findings (comment 26681 on #1508, pointer 26685 on #1507).
-Candidate: `agents/dewey/work/chat-03/I1-r2-manifest.sha256`, digest
-`2b48e333a0f09185364359ae6f8277cc88c0b9ff39058de45cc2c1f0ec9d5c4a`,
-the same 27 files, base 1c724958, uncommitted. Packet:
-`agents/dewey/work/chat-03/BUILD-I1-r2.md`.
-
-Verdict: approved for my part. B1 and B2 are fixed. Nothing must be fixed
-before the commit. Three follow-ups below, none of them blocking.
-
-## Checks
-
-- The manifest hashes to 2b48e333. All 27 files match it in the canonical
-  working tree. Twelve changed from round 1: README, `controller.mjs`,
-  `pi-pin.mjs`, `terminal.mjs` and seven test files. `engine.mjs` is
-  unchanged.
-- Export: `git archive` of 1c724958 plus the 27 files in
-  `~/darkwing-scratch/r5r2/exp`, manifest OK there too, `TMPDIR` under the
-  same directory.
-  - `node --test packages/conversation/tests/` passes 152/152, 0 skipped.
-  - `node --test packages/control-board/tests/` passes 124/124. The board
-    imports the conversation package.
-  - The CHAT-00 (48 checks), CHAT-01 R3 and CHAT-01c checks still pass.
-- Probe script: `~/darkwing-scratch/r5r2/probes.mjs`, output in
-  `~/darkwing-scratch/r5r2/out/probes.txt`. It drives the exported
-  `Controller` directly, as an outside caller would.
-
-## B1, the seal is now an allow-list: fixed
-
-`checkSeal` (pi-pin.mjs 68) requires the exact prefix `--mode rpc`, the
-seal flags and `--session <absolute path>`, then accepts only `--model`,
-`--provider` and `--thinking`, each once, each with one value that is
-nonempty and doesn't start with `-` or `@`. The constructor refuses a
-non-list `preArgs` or `extraArgs` and runs the seal before anything else
-happens (controller.mjs 170–174), so a refused argv never reaches a spawn.
-
-I passed 24 `extraArgs` lists to the constructor. These all refuse
-`unsealed-engine`:
-- my round 1 vectors: `--session <outside file>`, `--mode json`,
-  `--mode text`, `--no-session`, `--fork <file>`, `--export <file>`;
-- `--print`, `-p hi`, a bare word, `@<file>`, `--approve`,
-  `--no-extensions`, `--extension x`, `-e x`;
-- `--model=x`, `--model` with no value, an empty value, a value of `-p`
-  or `@f`, `--model` twice, and `--model a hello`.
-
-These build: `--model a --thinking high --provider p`, `--thinking high`,
-and `--model "rpc --mode json"`. The last one is a single argv element with
-no shell in between, so Pi sees it as a model name. That's fine.
-
-Dewey's mutants r2-B1 to r2-B1e (no extraArgs check, no prefix order, a
-relative session path, a repeat, a flag or `@` value) are all killed by
-N24.
-
-## B2, the session key is the Pi header ID: fixed
-
-The constructor reads the session header and keys the claim on
-`sessionKey({ harness: "pi", nativeSession })` (controller.mjs 193–194).
-I ran two controllers, seat A on the fixture session and seat B on a
-second file under another seat directory:
-
-| Second file | A | B | Keys equal |
-|---|---|---|---|
-| hard link of A's file | active, 1 launch | refused `already-active`, 0 launches | yes |
-| copy of A's file | active, 1 launch | refused `already-active`, 0 launches | yes |
-| copy with a different header ID | active, 1 launch | active, 1 launch | no |
-
-So the fix doesn't over-collide: two real sessions still start side by
-side. I also replaced the header ID by rename after construction. `start()`
-refused `target` with no launch. Mutants r2-B2 and r2-B2b are killed by
-the new W4 cases.
-
-n1 (Pi's startup-append rule) and n2 (the guard follows `$HOME`) are fixed
-in the README as asked.
-
-## What the rework touched that I checked
-
-- `#readSession` runs after the guard check at construction and wraps an
-  unreadable file as `configuration`. Good.
-- The force-stop path now refuses `fenced` while an escalation runs
-  (controller.mjs 683). `#forceStop` holds `escalating` and clears it in
-  `finally` (1453–1459). See F2 for the one gap.
-- `stopLink` needs `abortWritten`, and there's an overlap recheck after the
-  pause before the abort. Mutants r2-n1 and r2-n2 are killed. This is
-  outside my part, so I note it without a ruling.
-- Mutant r2-B5b (the shim writes the freeze but doesn't wait for
-  `frozen 1`) survives. That is Filbert's finding and Dewey explains it in
-  the packet. I leave it to Filbert.
-
-## Follow-ups (none must be fixed before the commit)
-
-- **F1. `engine.command` and `engine.preArgs` are outside the seal.** The
-  README documents them as a test hook, as I asked in round 1, and
-  `preArgs` still refuses `--extension`. But `preArgs:
-  [<pinned cli.js>, "--no-session"]` builds. Everything after `cli.js`
-  reaches Pi's parser, so a flag there that the seal doesn't repeat later,
-  such as `--no-session` or `--export`, takes effect. A non-default
-  `command` can run anything. No caller passes them today: the package
-  has no entry point that builds a `Controller` from configuration. Before I3 adds one, either
-  refuse any non-default `command`/`preArgs` outside the test harness, or
-  have that entry point never accept them from config. Owner: whoever
-  builds the I3 entry point.
-- **F2. `escalating` is set before `after()` runs.** The force-stop branch
-  of `#evaluate` sets `this.escalating` at controller.mjs 686, then calls
-  `#admission()` and `link.poison()`. If either throws, `#handle` turns the
-  result into an internal error and drops `after`. `#forceStop` never runs,
-  so the flag is never cleared, and every later force stop on that
-  controller refuses `fenced`. Recovery then needs a controller restart.
-  Neither call is expected to throw, so the risk is low. Fix: set the flag
-  only in `#forceStop`, or clear it in the catch path when `after` is
-  dropped.
-- **F3. `engine.env` defaults to `process.env`** (controller.mjs 168). A
-  real Pi then inherits the caller's whole environment, including
-  provider keys and the real `HOME`, so it reads the real `~/.pi/agent`.
-  This is unchanged from round 1 and fine for I1, where tests pass their
-  own env. The I3 entry point should build the engine env from an explicit
-  list.

agents/darkwing/work/cohort-k1/build-manifest.sha256

@@ -1 +0,0 @@
-fd10b62c10bff4736b9b4e809b283fe4c5b549fa53fe1121a7bb06258147f0e9  packages/conversation/tests/fake-pi.mjs

agents/darkwing/work/cohort-k1/build.md

@@ -1,152 +0,0 @@
-# Row 46: K1, K3 and K10 on the scope fixtures
-
-Darkwing, 2026-10-09. Issue #1528, reviewer Dewey. Brief:
-`docs/plans/2026-10-09_s4-follow-up-and-cohort.md`, section "Conversation
-cohort: K1, K3 and K10 fail on the scope fixtures". Ruling: lead decision 72.
-The candidate is not committed, staged or pushed.
-
-## Cause
-
-It's a race in the test fixture. Neither the host's systemd setup nor
-`cohort.mjs` is at fault.
-
-`spawnChild` in `packages/conversation/tests/fake-pi.mjs` returns the
-child's pid as soon as `spawn()` returns. The child is `node -e`, and it
-installs its SIGTERM handler only after Node has booted. That takes 16 to
-22 ms on an idle host (`trace-pass.jsonl`) and 43 to 80 ms under 48 CPU burners (`trace-load.jsonl`). A
-TERM that arrives before the handler gets the default action, and the child
-dies of SIGTERM. Each failing assertion is that death seen from a different
-place:
-
-- K1, "the escaped child is a listed member". The child died during the
-  TERM grace, so the freeze-phase enumeration doesn't list it.
-- K3, "the member ignored TERM". `alive(child)` is false at the kill
-  phase.
-- K10, "the member is alive across the crash". Same as K3, before the
-  restart.
-
-The time between spawn and TERM depends on the disk under TMPDIR. The
-fixture puts the claim store there. Before it sends TERM, the controller
-publishes claim revisions, each with an fsync on the file and one on the
-directory (`packages/conversation/src/claim.mjs:159` and `:176`).
-
-| TMPDIR | Disk | write+fsync median (`fsync.txt`) | spawn to TERM | Unfixed K1/K3/K10 |
-|---|---|---|---|---|
-| `/mnt/storage/scratch/tmp` | nvme0, ext4 | 0.65 ms | 12 to 13 ms (`trace-fail.jsonl`) | 0/3 (`runs/canon-scratchtmp.txt`) |
-| `/tmp` | nvme1, ext4 | 4.63 ms | not traced | 3/3 (`runs/canon-tmp.txt`) |
-| `~/darkwing-scratch/tmp` | nvme1, ext4 | 4.69 ms | 53 to 110 ms (`trace-pass.jsonl`) | 3/3 (`runs/canon-hometmp.txt`) |
-
-On the fast disk, TERM lands about 12 ms after spawn, before Node is up.
-In `trace-fail.jsonl` the children never log `ready`.
-
-### Why it started failing
-
-This host's seats now get `TMPDIR=/mnt/storage/scratch/tmp`. My shell had
-it by default, and so did Sage's gate runs. It comes from Vikunja task 59.
-`~/.config/systemd/user/t3code.service.d/tmpdir.conf` was written
-2026-10-04 20:33Z. Its own comment says it takes effect only when
-`t3code.service` restarts, which hasn't happened (active since 2026-09-23),
-and that until then seats get TMPDIR from their harness config. I didn't
-establish what TMPDIR the row 44 gate ran with on 2026-10-05, so the
-"since when" is likely but unproven. The change that exposed the race is a
-seat's TMPDIR. It isn't a systemd or user manager setting, and I changed
-nothing on the host.
-
-### The scope is not a factor
-
-Outside any scope, `race.mjs` gives the same split: a TERM 0, 10 or 20 ms
-after spawn kills 10/10, and at 30 or 60 ms 10/10 survive (`race.txt`).
-Sage's outside-scope probe found the child surviving. I haven't seen that
-probe, but it most likely sent TERM after the handler was in place.
-
-## Receipts in a scope
-
-`scope-receipt.mjs` launches a delegated scope per run, with the same
-`systemd-run` flags `ScopeLauncher` uses. The scope's main process spawns
-the fixture's `ignoreTerm` child and records `systemctl --user show` on the
-scope, then `cgroup.procs` before and after TERM, then the child's exit.
-Output: `receipts.jsonl`.
-
-| Mode | TERM after spawn | In `cgroup.procs` before | After | Child exit |
-|---|---|---|---|---|
-| `race 12` | 20 to 25 ms | 5/5 | 1/5 | 4/5 `signal: "SIGTERM"`, 1 alive |
-| `race 60` | 69 to 71 ms | 5/5 | 5/5 | 5/5 alive 500 ms after TERM |
-| `ready 0` | 28 to 38 ms (ready at 22 to 30) | 5/5 | 5/5 | 5/5 alive 500 ms after TERM |
-
-Each line also carries the scope's `Id`, `LoadState=loaded`,
-`ActiveState=active`, `InvocationID`, `ControlGroup` and `Delegate=yes`.
-The first line, for example:
-`ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-12-3662746-0.scope`,
-`before [3662760, 3662788]`, `after [3662760]`, child exit
-`{"code":null,"signal":"SIGTERM","atMs":24}`.
-
-`trace.patch` is the scratch-only instrumentation behind the two traces.
-It logs spawn, ready and the shim's `term` per pid to `$DW_TRACE`. It is
-not part of the candidate.
-
-## The fix
-
-`build.patch` changes one file, `packages/conversation/tests/fake-pi.mjs`:
-
-- The child writes one byte to stdout as its first action after installing
-  its TERM handler. A child without `ignoreTerm` writes it once its code
-  starts.
-- `spawnChild` returns a promise. It resolves with the pid on that byte,
-  then closes its end of the pipe. It rejects if the child exits or errors
-  first, or if 10 s pass.
-- The `child` control op returns that promise. The op dispatcher now
-  answers `ok: false` when an op's promise rejects. Before, a rejected op
-  promise went unhandled, and under Node's default that ends fake-pi. The
-  change covers every op that returns a promise, not only `child`.
-  `childOf` in `cohort.test.mjs` already asserts `r.ok`.
-
-No assertion changed, and `cohort.mjs` is untouched. The three cases now
-test what their names say: TERM reaches a child that is already ignoring
-TERM.
-
-`build-manifest.sha256` (sha256
-`9228414352a56e0465e896b81643cdce000bcedba70f86adac07fc50cfadad2d`) pins
-`fake-pi.mjs` after the patch. `build.patch` sha256 is
-`04234ce1d886e36dfd06cbd81153b38cbb86909b6ee933151306b8cdedd629d9`. In a
-fresh worktree at `521597bb` the patch applies and the manifest checks 1/1.
-
-## Mutants
-
-Both ran with `TMPDIR=/mnt/storage/scratch/tmp`, the condition that fails.
-
-| Mutant | K1/K3/K10 | File |
-|---|---|---|
-| M1: resolve at spawn, no wait (the old behavior) | 0/3 | `runs/mutant-m1-nowait.txt` |
-| M2: wait for ready, but the child has no TERM handler | 0/3, the same three assertions | `runs/mutant-m2-noignore.txt` |
-
-M1 shows the wait is what fixes it. M2 shows the assertions still catch a
-child that dies on TERM.
-
-## Runs
-
-The patch was applied in a scratch worktree at `521597bb`. Node v26.8.1.
-Start time and load are in `runs/start.txt`.
-
-| Run | TMPDIR | Result | File |
-|---|---|---|---|
-| `node --test "packages/conversation/tests/*.test.mjs"` | `/mnt/storage/scratch/tmp` | 152/152 | `runs/node-conversation.txt` |
-| `node --test "packages/webui/tests/*.test.mjs"` | `/mnt/storage/scratch/tmp` | 14/14 | `runs/node-webui.txt` |
-| K1/K3/K10 isolated, 3 consecutive | `/mnt/storage/scratch/tmp` | 3/3, 3/3, 3/3 | `runs/k-iso-{1,2,3}.txt` |
-| K1/K3/K10 isolated | `/tmp` | 3/3 | `runs/k-tmp.txt` |
-| K1/K3/K10 isolated | `~/darkwing-scratch/tmp` | 3/3 | `runs/k-home.txt` |
-| K1/K3/K10 isolated under 48 CPU burners, 3 runs (load 13.6 to 24.9) | `/mnt/storage/scratch/tmp` | 3/3, 3/3, 3/3 | `runs/k-load48-{1,2,3}.txt` |
-
-For the gate rerun, keep the default `TMPDIR=/mnt/storage/scratch/tmp`.
-That's the condition that failed, and a slower TMPDIR would pass with or
-without the patch.
-
-## Files
-
-- `build.md`, this file.
-- `build.patch`, `build-manifest.sha256`: the candidate.
-- `scope-receipt.mjs`, `receipts.jsonl`: the scope receipts.
-- `race.mjs`, `race.txt`: TERM timing outside a scope.
-- `fsync.mjs`, `fsync.txt`: write+fsync latency per TMPDIR.
-- `trace.patch`, `trace-fail.jsonl`, `trace-pass.jsonl`, `trace-load.jsonl`: the traced runs.
-- `runs/`: suite, isolated, load and mutant outputs, and the unfixed
-  canonical tree under three TMPDIRs.

agents/darkwing/work/cohort-k1/build.patch

@@ -1,67 +0,0 @@
-diff --git a/packages/conversation/tests/fake-pi.mjs b/packages/conversation/tests/fake-pi.mjs
-index b3014f86..a78dfebd 100644
---- a/packages/conversation/tests/fake-pi.mjs
-+++ b/packages/conversation/tests/fake-pi.mjs
-@@ -615,14 +615,34 @@ const children = [];
- // K2); `forkLoop` forks every 5 ms (K12); `ignoreTerm` survives SIGTERM, so
- // only the kill phase ends it (K3, K10, K11). With `pidLog`, the fork loop
- // appends each child's pid and a `term` line when it gets SIGTERM (K12).
-+//
-+// It resolves only once the child writes its ready byte, which it does after
-+// installing its TERM handler. Node takes 15 to 80 ms to get there, and a
-+// force stop can send TERM sooner (12 ms when the claim store is on a fast
-+// disk). A TERM before the handler kills a child that is meant to ignore it
-+// (#1528).
- function spawnChild({ setsid = false, forkLoop = false, ignoreTerm = false, pidLog = null } = {}) {
-   const note = pidLog ? `const note=(s)=>require('node:fs').appendFileSync(${JSON.stringify(pidLog)},s+'\\n');` : "const note=()=>{};";
--  const code = note + (ignoreTerm ? "process.on('SIGTERM',()=>note('term'));" : "") + (forkLoop
-+  const code = note + (ignoreTerm ? "process.on('SIGTERM',()=>note('term'));" : "") + "process.stdout.write('r');" + (forkLoop
-     ? "const {spawn}=require('node:child_process');setInterval(()=>{try{const c=spawn('sleep',['1000'],{stdio:'ignore'});if(c.pid)note(String(c.pid))}catch{}},5);setInterval(()=>{},1e9)"
-     : "setInterval(()=>{},1e9)");
--  const child = spawn(process.execPath, ["-e", code], { stdio: "ignore", detached: setsid });
-+  const child = spawn(process.execPath, ["-e", code], { stdio: ["ignore", "pipe", "ignore"], detached: setsid });
-   children.push(child.pid);
--  return child.pid;
-+  return new Promise((resolve, reject) => {
-+    const fail = (why) => {
-+      clearTimeout(timer);
-+      reject(new Error(`tool child ${child.pid} ${why} before it was ready`));
-+    };
-+    const timer = setTimeout(() => fail("took 10 s"), 10000);
-+    child.once("error", (err) => fail(err.code ?? err.message));
-+    child.once("exit", (code, signal) => fail(`exited (${signal ?? code})`));
-+    child.stdout.once("data", () => {
-+      clearTimeout(timer);
-+      child.removeAllListeners("exit");
-+      child.stdout.destroy();
-+      resolve(child.pid);
-+    });
-+  });
- }
- 
- // K13: a member writes its own pid to another cgroup's cgroup.procs.
-@@ -671,7 +691,7 @@ async function main() {
-         drop: () => fake.dropResponse(req.type, req.n ?? 1),
-         extension: () => void fake.extensionPrompt(req.args ?? {}),
-         state: () => ({ streaming: fake.streaming, runs: fake.runs.length, commands: fake.commands, pid: process.pid, children, appends: fake.appends }),
--        child: () => ({ pid: spawnChild(req.args ?? {}) }),
-+        child: () => spawnChild(req.args ?? {}).then((pid) => ({ pid })),
-         escape: () => escape(req.target),
-         cgroup: () => readFileSync(`/proc/${req.pid ?? process.pid}/cgroup`, "utf8"),
-         waitPaused: () => fake.waitPaused(req.point),
-@@ -679,12 +699,13 @@ async function main() {
-         stall: () => void process.stdin.pause(),
-       };
-       if (!ops[req.op]) return sock.write(encodeLine({ id: req.id, ok: false, error: `unknown op ${req.op}` }));
-+      const failed = (err) => sock.write(encodeLine({ id: req.id, ok: false, error: String(err.message) }));
-       try {
-         const out = ops[req.op]();
--        if (out && typeof out.then === "function") out.then(reply);
-+        if (out && typeof out.then === "function") out.then(reply, failed);
-         else reply(out ?? null);
-       } catch (err) {
--        sock.write(encodeLine({ id: req.id, ok: false, error: String(err.message) }));
-+        failed(err);
-       }
-     });
-     sock.on("data", (c) => splitter.push(c));

agents/darkwing/work/cohort-k1/fsync.mjs

@@ -1,17 +0,0 @@
-import { openSync, writeSync, fsyncSync, closeSync, rmSync, mkdtempSync } from "node:fs";
-import { join } from "node:path";
-for (const base of process.argv.slice(2)) {
-  const d = mkdtempSync(join(base, "dw-fsync-"));
-  const t = [];
-  for (let i = 0; i < 30; i++) {
-    const s = performance.now();
-    const fd = openSync(join(d, `f${i}`), "w");
-    writeSync(fd, "x".repeat(512));
-    fsyncSync(fd);
-    closeSync(fd);
-    t.push(performance.now() - s);
-  }
-  rmSync(d, { recursive: true });
-  t.sort((a, b) => a - b);
-  console.log(`${base}: write+fsync median ${t[15].toFixed(2)} ms, p90 ${t[27].toFixed(2)} ms`);
-}

agents/darkwing/work/cohort-k1/fsync.txt

@@ -1,3 +0,0 @@
-/mnt/storage/scratch/tmp: write+fsync median 0.65 ms, p90 0.79 ms
-/tmp: write+fsync median 4.63 ms, p90 4.93 ms
-/home/jwoltje/darkwing-scratch/tmp: write+fsync median 4.69 ms, p90 9.30 ms

agents/darkwing/work/cohort-k1/race.mjs

@@ -1,17 +0,0 @@
-// Spawn the K fixtures' ignoreTerm child exactly as fake-pi.mjs spawnChild does, then
-// SIGTERM it after a delay. Reports how the child ended within 1 s.
-import { spawn } from "node:child_process";
-const code = "const note=()=>{};process.on('SIGTERM',()=>note('term'));setInterval(()=>{},1e9)";
-const delays = process.argv.slice(2).map(Number);
-for (const d of delays) {
-  const r = { died: 0, survived: 0 };
-  for (let i = 0; i < 10; i++) {
-    const c = spawn(process.execPath, ["-e", code], { stdio: "ignore" });
-    const ended = new Promise((res) => c.on("exit", (code, sig) => res(sig ?? code)));
-    await new Promise((res) => setTimeout(res, d));
-    c.kill("SIGTERM");
-    const out = await Promise.race([ended, new Promise((res) => setTimeout(() => res(null), 1000))]);
-    if (out === null) { r.survived++; c.kill("SIGKILL"); await ended; } else r.died++;
-  }
-  console.log(`TERM ${d} ms after spawn: died ${r.died}/10 survived ${r.survived}/10`);
-}

agents/darkwing/work/cohort-k1/race.txt

@@ -1,5 +0,0 @@
-TERM 0 ms after spawn: died 10/10 survived 0/10
-TERM 10 ms after spawn: died 10/10 survived 0/10
-TERM 20 ms after spawn: died 10/10 survived 0/10
-TERM 30 ms after spawn: died 0/10 survived 10/10
-TERM 60 ms after spawn: died 0/10 survived 10/10

agents/darkwing/work/cohort-k1/receipts.jsonl

@@ -1,15 +0,0 @@
-{"mode":"race","delayMs":12,"readyMs":null,"termMs":23,"child":3662788,"self":3662760,"show":["Id=dw-r46-race-12-3662746-0.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=2eac4fa367504b9dbca66a4693cc328a","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-12-3662746-0.scope","Delegate=yes"],"before":[3662760,3662788],"after":[3662760],"childInBefore":true,"childInAfter":false,"exit":{"code":null,"signal":"SIGTERM","atMs":24}}
-{"mode":"race","delayMs":12,"readyMs":null,"termMs":20,"child":3663479,"self":3663212,"show":["Id=dw-r46-race-12-3662746-1.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=021c6d466fc64961ba3031a0254c3803","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-12-3662746-1.scope","Delegate=yes"],"before":[3663212,3663479],"after":[3663212],"childInBefore":true,"childInAfter":false,"exit":{"code":null,"signal":"SIGTERM","atMs":21}}
-{"mode":"race","delayMs":12,"readyMs":null,"termMs":25,"child":3663799,"self":3663588,"show":["Id=dw-r46-race-12-3662746-2.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=ce8e2c2dbbe7481885a50152fe3766b5","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-12-3662746-2.scope","Delegate=yes"],"before":[3663588,3663799],"after":[3663588],"childInBefore":true,"childInAfter":false,"exit":{"code":null,"signal":"SIGTERM","atMs":27}}
-{"mode":"race","delayMs":12,"readyMs":null,"termMs":20,"child":3664563,"self":3664322,"show":["Id=dw-r46-race-12-3662746-3.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=cc912ab9a7fe47dd9e3b97486abf0bc3","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-12-3662746-3.scope","Delegate=yes"],"before":[3664322,3664563],"after":[3664322,3664563],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"race","delayMs":12,"readyMs":null,"termMs":21,"child":3665177,"self":3665003,"show":["Id=dw-r46-race-12-3662746-4.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=163271c658cc4ee4b84f5aace54bc806","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-12-3662746-4.scope","Delegate=yes"],"before":[3665003,3665177],"after":[3665003],"childInBefore":true,"childInAfter":false,"exit":{"code":null,"signal":"SIGTERM","atMs":22}}
-{"mode":"race","delayMs":60,"readyMs":null,"termMs":69,"child":3665812,"self":3665626,"show":["Id=dw-r46-race-60-3665599-0.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=14681de5ac844879bdfcbcdd3dae8f10","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-60-3665599-0.scope","Delegate=yes"],"before":[3665626,3665812],"after":[3665626,3665812],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"race","delayMs":60,"readyMs":null,"termMs":71,"child":3666377,"self":3666179,"show":["Id=dw-r46-race-60-3665599-1.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=3128eba3ed234776ae4a61cb12fe1210","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-60-3665599-1.scope","Delegate=yes"],"before":[3666179,3666377],"after":[3666179,3666377],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"race","delayMs":60,"readyMs":null,"termMs":70,"child":3666924,"self":3666728,"show":["Id=dw-r46-race-60-3665599-2.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=836de499898d4b6996739e1b2a433fa0","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-60-3665599-2.scope","Delegate=yes"],"before":[3666728,3666924],"after":[3666728,3666924],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"race","delayMs":60,"readyMs":null,"termMs":69,"child":3667387,"self":3667251,"show":["Id=dw-r46-race-60-3665599-3.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=0ebf8d9d930a43629668e10c08f80e0d","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-60-3665599-3.scope","Delegate=yes"],"before":[3667251,3667387],"after":[3667251,3667387],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"race","delayMs":60,"readyMs":null,"termMs":70,"child":3667888,"self":3667727,"show":["Id=dw-r46-race-60-3665599-4.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=9c3594ae3018458594bb3ff996dc0d9a","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-race-60-3665599-4.scope","Delegate=yes"],"before":[3667727,3667888],"after":[3667727,3667888],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"ready","delayMs":0,"readyMs":30,"termMs":38,"child":3668171,"self":3668085,"show":["Id=dw-r46-ready-0-3668072-0.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=4c2c3b3ad52a477e8f735abedf158cbb","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-ready-0-3668072-0.scope","Delegate=yes"],"before":[3668085,3668171],"after":[3668085,3668171],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"ready","delayMs":0,"readyMs":22,"termMs":29,"child":3668354,"self":3668284,"show":["Id=dw-r46-ready-0-3668072-1.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=b9da06280a8a4c6682f05a2d4ff118cf","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-ready-0-3668072-1.scope","Delegate=yes"],"before":[3668284,3668354],"after":[3668284,3668354],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"ready","delayMs":0,"readyMs":22,"termMs":28,"child":3668425,"self":3668417,"show":["Id=dw-r46-ready-0-3668072-2.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=aab0b83d085e4bac86fdbe1ce65f23dc","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-ready-0-3668072-2.scope","Delegate=yes"],"before":[3668417,3668425],"after":[3668417,3668425],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"ready","delayMs":0,"readyMs":23,"termMs":30,"child":3668509,"self":3668482,"show":["Id=dw-r46-ready-0-3668072-3.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=d076a3b629554b83bded41cb00e667b8","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-ready-0-3668072-3.scope","Delegate=yes"],"before":[3668482,3668509],"after":[3668482,3668509],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}
-{"mode":"ready","delayMs":0,"readyMs":23,"termMs":30,"child":3668635,"self":3668624,"show":["Id=dw-r46-ready-0-3668072-4.scope","LoadState=loaded","ActiveState=active","SubState=running","InvocationID=6b46a0b279ca454eaa034d48a4de7385","ControlGroup=/user.slice/user-1000.slice/[email protected]/app.slice/dw-r46-ready-0-3668072-4.scope","Delegate=yes"],"before":[3668624,3668635],"after":[3668624,3668635],"childInBefore":true,"childInAfter":true,"exit":"alive 500 ms after TERM"}

agents/darkwing/work/cohort-k1/runs/canon-hometmp.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2737.880671ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2557.042479ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (861.841994ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 6282.04986

agents/darkwing/work/cohort-k1/runs/canon-scratchtmp.txt

@@ -1,56 +0,0 @@
-✖ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2537.385038ms)
-✖ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2331.351365ms)
-✖ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (209.932151ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 0
-ℹ fail 3
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 5448.48373
-
-✖ failing tests:
-
-test at packages/conversation/tests/cohort.test.mjs:139:1
-✖ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2537.385038ms)
-  AssertionError [ERR_ASSERTION]: the escaped child is a listed member
-      at TestContext.<anonymous> (file:///mnt/storage/src/mosaic-stack/packages/conversation/tests/cohort.test.mjs:151:12)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async startSubtestAfterBootstrap (node:internal/test_runner/harness:387:3) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }
-
-test at packages/conversation/tests/cohort.test.mjs:176:1
-✖ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2331.351365ms)
-  AssertionError [ERR_ASSERTION]: the member ignored TERM
-      at TestContext.<anonymous> (file:///mnt/storage/src/mosaic-stack/packages/conversation/tests/cohort.test.mjs:190:12)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }
-
-test at packages/conversation/tests/cohort.test.mjs:426:3
-✖ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (209.932151ms)
-  AssertionError [ERR_ASSERTION]: the member is alive across the crash
-      at TestContext.<anonymous> (file:///mnt/storage/src/mosaic-stack/packages/conversation/tests/cohort.test.mjs:431:12)
-      at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }

agents/darkwing/work/cohort-k1/runs/canon-tmp.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2360.366213ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2364.75763ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (628.04725ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 5474.064409

agents/darkwing/work/cohort-k1/runs/k-home.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2457.36226ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2375.081486ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (636.350971ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 5607.100296

agents/darkwing/work/cohort-k1/runs/k-iso-1.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2665.443725ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2531.922418ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (706.098022ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 6068.030335

agents/darkwing/work/cohort-k1/runs/k-iso-2.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2699.60543ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2492.169051ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (675.081525ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 6026.119198

agents/darkwing/work/cohort-k1/runs/k-iso-3.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2762.539396ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2528.214494ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (644.633095ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 6115.728995

agents/darkwing/work/cohort-k1/runs/k-load48-1.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2887.135899ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2915.708256ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (2590.181146ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 9662.912642

agents/darkwing/work/cohort-k1/runs/k-load48-2.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (3084.73509ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2750.8295ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (2151.224665ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 8355.450647

agents/darkwing/work/cohort-k1/runs/k-load48-3.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2746.07054ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2559.848734ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (1985.055632ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 7722.015242

agents/darkwing/work/cohort-k1/runs/k-tmp.txt

@@ -1,11 +0,0 @@
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2785.421246ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2619.007617ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (889.082424ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 3
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 6776.303473

agents/darkwing/work/cohort-k1/runs/mutant-m1-nowait.txt

@@ -1,56 +0,0 @@
-✖ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2523.134497ms)
-✖ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2421.121319ms)
-✖ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (211.187938ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 0
-ℹ fail 3
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 5461.571354
-
-✖ failing tests:
-
-test at cohort.test.mjs:139:1
-✖ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2523.134497ms)
-  AssertionError [ERR_ASSERTION]: the escaped child is a listed member
-      at TestContext.<anonymous> (file:///home/jwoltje/darkwing-scratch/r46/fix/packages/conversation/tests/cohort.test.mjs:151:12)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async startSubtestAfterBootstrap (node:internal/test_runner/harness:387:3) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }
-
-test at cohort.test.mjs:176:1
-✖ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2421.121319ms)
-  AssertionError [ERR_ASSERTION]: the member ignored TERM
-      at TestContext.<anonymous> (file:///home/jwoltje/darkwing-scratch/r46/fix/packages/conversation/tests/cohort.test.mjs:190:12)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }
-
-test at cohort.test.mjs:426:3
-✖ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (211.187938ms)
-  AssertionError [ERR_ASSERTION]: the member is alive across the crash
-      at TestContext.<anonymous> (file:///home/jwoltje/darkwing-scratch/r46/fix/packages/conversation/tests/cohort.test.mjs:431:12)
-      at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }

agents/darkwing/work/cohort-k1/runs/mutant-m2-noignore.txt

@@ -1,56 +0,0 @@
-✖ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2545.15719ms)
-✖ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2407.231229ms)
-✖ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (304.986006ms)
-ℹ tests 3
-ℹ suites 0
-ℹ pass 0
-ℹ fail 3
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 5605.846956
-
-✖ failing tests:
-
-test at cohort.test.mjs:139:1
-✖ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2545.15719ms)
-  AssertionError [ERR_ASSERTION]: the escaped child is a listed member
-      at TestContext.<anonymous> (file:///home/jwoltje/darkwing-scratch/r46/fix/packages/conversation/tests/cohort.test.mjs:151:12)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async startSubtestAfterBootstrap (node:internal/test_runner/harness:387:3) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }
-
-test at cohort.test.mjs:176:1
-✖ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2407.231229ms)
-  AssertionError [ERR_ASSERTION]: the member ignored TERM
-      at TestContext.<anonymous> (file:///home/jwoltje/darkwing-scratch/r46/fix/packages/conversation/tests/cohort.test.mjs:190:12)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }
-
-test at cohort.test.mjs:426:3
-✖ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (304.986006ms)
-  AssertionError [ERR_ASSERTION]: the member is alive across the crash
-      at TestContext.<anonymous> (file:///home/jwoltje/darkwing-scratch/r46/fix/packages/conversation/tests/cohort.test.mjs:431:12)
-      at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
-      at async Test.run (node:internal/test_runner/test:1409:7)
-      at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) {
-    generatedMessage: false,
-    code: 'ERR_ASSERTION',
-    actual: false,
-    expected: true,
-    operator: '==',
-    diff: 'simple'
-  }

agents/darkwing/work/cohort-k1/runs/node-conversation.txt

@@ -1,160 +0,0 @@
-✔ W1: two processes acquire the same pair at once; exactly one claim (124.186492ms)
-✔ W1: two writers publish the same revision at once: one wins, the other gets null, the winner's record stays (11.494379ms)
-✔ W1: a revision name appears only after its bytes are synced; before that, only a temp file exists (5.019078ms)
-✔ W2: acquire while a claim is reserved or active refuses already-active (169.129732ms)
-✔ W3: acquire while stopping, uncertain, or stopped without proof refuses unsafe-replacement (231.756886ms)
-✔ W4: same session with another seat tuple, and the reverse, both refuse; a loser on the seat key closes it no-unit (192.859955ms)
-✔ W4: a hard link of one session under another seat is the same session: the second controller refuses already-active and launches nothing (35.003128ms)
-✔ W4: a copy of one session under another seat is the same session: the second controller refuses already-active and launches nothing (21.132326ms)
-✔ W4: a session header ID that changes after construction refuses target; nothing is claimed or launched (3.242382ms)
-✔ W5: SIGKILL between every publication barrier of acquire and transition; restart never finds two holders or a lost claim (5566.910492ms)
-✔ W5: SIGKILL between every publication barrier of release; restart finishes or holds the release (22659.492505ms)
-✔ W6: controller killed mid-turn while the engine lives; restart is uncertain, no launch, prompts refuse (201.806859ms)
-✔ W12: a live owner paused with SIGSTOP; a second controller refuses already-active and changes nothing (116.242029ms)
-✔ W13: crash after the engine spawns, before active; restart finds the live unit: uncertain, no second spawn, force stop only (262.226787ms)
-✔ W14: crash after reservation, before the spawn marker: stopped with a no-unit observation; the pair is free (220.425841ms)
-✔ W20: crash after the spawn marker, scope collected; uncertain in both runs, the marker is copied, no launch until a boot proof (240.127173ms)
-✔ W15: crash between the two keys during release; restart finishes it under the same claim ID (35.945246ms)
-✔ W7: recorded boot ID differs on the same machine: stopped with a boot proof; open tool calls become uncertain (98.424488ms)
-✔ W8: resume after a proven stop with the same pins: new claim ID, generation +1, same conversation, branch and leaf (32.141383ms)
-✔ W9: resume with a changed binary, argv digest, branch or leaf is refused and the claim is unchanged (89.672997ms)
-✔ W11: the controller writes no session file; only the fake engine's own appends appear (23.121134ms)
-✔ W16: a highest revision that won't parse holds the pair uncertain; the older stopped revision is not reused (54.682806ms)
-✔ W17: a claim root copied from another host refuses foreign-host and promotes nothing (57.719055ms)
-✔ G1: a session path or claim root under .pi/state, ~/.claude, the data root or a registration refuses at construction (4.12463ms)
-✔ G2: a symlink inside the fixture root to a live session file is refused by the real-path check (1.12384ms)
-✔ G3: a fixture path swapped for a live path after construction is refused at bind (1.84306ms)
-✔ K1: force stop kills a tool child that called setsid; stopped with a verified proof (2507.919568ms)
-✔ K2: K1 on the process-group fallback ends uncertain, never stopped (153.327228ms)
-✔ K3: SIGTERM acknowledged while a member lives: stopping until the kill phase, never stopped from TERM (2414.676654ms)
-✔ K4: two engines; force stop one; the other survives by independent observation (4314.515513ms)
-✔ K5: a stop during a tool call leaves the effect uncertain, and it is shown (2190.378919ms)
-✔ K12: a member forking in a loop: the freeze stops it, enumeration is complete, populated 0 after cgroup.kill (2244.796524ms)
-✔ K13: a member writing its pid into another cgroup is refused by the namespace; the kill is complete (2189.63935ms)
-✔ K15: the shim gone, engine/cgroup.events unreadable, or the engine cgroup missing: evidence unavailable, not empty; uncertain (4473.315326ms)
-✔ K10: controller killed between the TERM and kill phases: restart checks the invocation ID and re-runs from TERM for the same stop (482.430287ms)
-✔ K11: controller killed after the confirmation is recorded, before TERM: restart checks the invocation ID and re-runs from TERM for the same stop (408.721794ms)
-✔ K14: a unit with the recorded name but another invocation ID: evidence unavailable, no signals, uncertain (307.075836ms)
-✔ K6: recover without proof, without confirmation, or with changed pins is refused (71.309069ms)
-✔ K7: recover after proof, then launch: new claim and execution, generation +1, same leaf; the cancelled prompt is not replayed (35.403696ms)
-✔ K8: an engine that loads another leaf on resume is refused before admission; it stays claimed until a proven stop (45.424183ms)
-✔ K9: an interrupt that never settles stays uncertain; force stop stays available; takeover is refused while fenced (3030.941578ms)
-✔ K16: a claim from another machine ID refuses foreign-host; no boot proof is issued (5.325961ms)
-✔ K17: two launcher calls with one eligibility record: one launch, the other refuses, no second engine (30.688596ms)
-✔ K18: the leaf changes after eligibility: launch refused; the reservation stays until released with proof (24.893734ms)
-✔ S1: `/goal x`, with leading spaces or a tab, refuses text-policy at admission; zero engine bytes (34.176579ms)
-✔ S2: every prefix pinned Pi interprets is refused, from the list the code uses; the rest reach the engine exactly (30.425031ms)
-✔ S3: `/goal` on the second line is pinned from the source: Pi checks only index 0, so it is admitted and sent exactly (27.293519ms)
-✔ S4: a `/` left in the composer is cleared when control transfers and returns; the next submit sends only the new text (42.046162ms)
-✔ S5: an observer terminal gets a paste then Enter, as send-message.sh does: not admitted: controller, nothing sent (21.340651ms)
-✔ S6: a mediated-shaped registration (no tmux) passed to the board's replyToRow: 409 no tmux session; exec never runs (1.34914ms)
-✔ S7: ESC, bracketed-paste markers and U+2028/U+2029 travel as one JSON string; the engine receives the exact text in one record (29.201111ms)
-✔ P3: a Pi confirm, select, input or editor dialog is shown disabled with a reason and never answered (138.365864ms)
-✔ E1: send, ack, user, toolCall, toolResult, final answer: shown once, no refresh, draft and reading position kept (39.366805ms)
-✔ E2: U+2028, U+2029 inside JSON strings and CRLF line ends each parse as one record, on the splitter and through the controller (21.443107ms)
-✔ E3: a multipart final, two blocks, null request correlation and duplicate delivery (31.257304ms)
-✔ E4: a page read after message_end but before its entry is persisted: marker at the seam, re-read after run-settled, each message once (26.594189ms)
-✔ E4: a gap or a new epoch also reconciles; nothing is concatenated across a gap (11.181347ms)
-✔ E5: an unknown native event gives no client event; evidence records its type and bytes; the terminal count goes up (26.220793ms)
-✔ E6: a tool result delayed across a pause and a reconnect is reconciled without a manual refresh (42.195378ms)
-✔ E7: the terminal renders the same stream as the library client, as observer and then as controller, and submits only as controller (41.055698ms)
-✔ terminal: engine control characters are made visible; a lost connection refuses submit (28.737455ms)
-✔ terminal: outcome unknown is shown as such, with no resend offer, and nothing is resent (0.53917ms)
-✔ terminal: text after Enter in the same input chunk starts the next message; it never joins the one submitted (0.351376ms)
-✔ terminal: a paste-start marker split right after its ESC still opens the paste; the Enter inside it never submits (0.356358ms)
-✔ terminal: invisible and bidi characters are made visible; head, status and notice lines stay one line (0.168609ms)
-✔ every record these fixtures produced is a valid CHAT-01 record (E5: no record fails the schema) (359.398761ms)
-✔ H1: two takeovers with the same expected generation: one wins, +1; the other refuses generation (56.47649ms)
-✔ H2: the old controller's prompt after a takeover commits is refused with zero engine bytes (79.112044ms)
-✔ H3: a takeover while a prompt holds the dispatch lock: written under the old actor, or refused; never both (131.150736ms)
-✔ H4: self-takeover is refused (22.440288ms)
-✔ H9: Interrupt racing a prompt's dispatch: before the write, dispatch-refused and no-turn; after, §3 rules (87.314186ms)
-✔ H10: Interrupt and force stop together: one stop chain, force stop supersedes (104.583327ms)
-✔ H10: an overlap during the pause before the abort: no abort, the stop ends uncertain (35.050855ms)
-✔ H10: a no-turn Interrupt lifts only its own fence; admission stays closed under force stop, overlap or revocation (70.09079ms)
-✔ H11: the controller disconnects mid-turn: work continues, the claim is unchanged, control stays put (129.262166ms)
-✔ H12: an exact retry after reconnecting to the same incarnation returns the same receipt; one dispatch (15.147609ms)
-✔ H13: a retry with the same request ID and different text is refused (13.200342ms)
-✔ H14: late stdout from the old engine after a replacement is dropped by incarnation, counted, never rendered (133.123623ms)
-✔ H15: a revoked connection's command is refused; the revocation fence holds (65.378107ms)
-✔ H16: a second controller for the same session refuses already-active; the first is untouched (16.417034ms)
-✔ H10: a second force stop while the first escalation runs refuses fenced; one escalation, and the claim records only the first stop's phases (64.083554ms)
-✔ H17: a confirmation reused, answered from another connection, or used after the stop changed is refused (56.969236ms)
-✔ H18: two prompts before any native output: the second refuses busy; one engine write (15.033626ms)
-✔ H19: the pipe fails mid-line under a large prompt: delivery-unknown transport-unknown, poisoned, no later write (119.397422ms)
-✔ H19: the link itself never writes again after an unknown outcome, whoever calls it (0.606652ms)
-✔ H19: the controller dies mid-write of a large line: after restart the outcome is unknown and nothing is resent (464.542066ms)
-✔ H20: the line is written but the ack is lost when the controller dies: orphan, outcome unknown, nothing resent (357.812591ms)
-✔ H21: a retry of the exact request with the old token after a crash is stale-incarnation; no second write (337.123938ms)
-✔ H22: after H21 and a valid recovery, a new request with the new token is admitted (2377.836435ms)
-✔ H23: requests pending at a restart are not resent; each shows outcome unknown (482.017907ms)
-✔ a plain conversation: catalogue row, one page, CHAT-01 records (10.196509ms)
-✔ native entries map to blocks: tools, thinking, bash, notices, ids that do not fit (3.091521ms)
-✔ F1: a malformed line is an unavailable part at its position, and reading continues (2.999666ms)
-✔ F1: a missing parent stops the history with a notice that names the unreadable lines (3.746417ms)
-✔ F1: an unreadable fork is never merged into another branch's history (3.733154ms)
-✔ F1: a follow stays on its branch when the next entry's parent is unreadable (3.881773ms)
-✔ F1: a file whose entries are all unreadable shows a notice per line (1.563335ms)
-✔ F2: a truncated trailing line marks the view incomplete, not an error (2.494119ms)
-✔ pagination: 100 parts, then the rest; parts concatenate to the whole branch (5.109673ms)
-✔ F3: a replaced file (new inode) refuses old cursors with reconcile (4.939351ms)
-✔ F4: a same-inode rewrite of the prefix refuses old cursors with reconcile (5.516645ms)
-✔ F5: growth between pages keeps the epoch and the page stops at the pinned length (5.793786ms)
-✔ F6: unknown, foreign and expired cursors refuse and leave the cursor usable (10.497498ms)
-✔ F7: a symlinked file and a symlinked directory component are refused, never opened (9.313922ms)
-✔ F8: a file swapped for a symlink after the catalogue is refused (2.247728ms)
-✔ F9: registrations never add or redirect a root (2.060491ms)
-✔ F10: a header cwd naming another project is refused (3.630852ms)
-✔ F11: parentSession renders with a marker and the parent is never opened (0.936984ms)
-✔ F12: two leaves: the default leaf is shown and the other branch reads alone (4.789444ms)
-✔ F12: a follow refuses when an appended duplicate id changes the branch's earlier parts (2.415895ms)
-✔ F12: a second root (Pi's resetLeaf) starts its own branch (1.349225ms)
-✔ F13: compaction is a marker in place, then the retained content (0.756245ms)
-✔ F14: long strings split into fragments and parts, reassemble exactly, and pages respect the byte cap (737.947726ms)
-✔ fragments never cut a surrogate pair and keep an empty string (9.768434ms)
-✔ F15: a Claude seat is an unsupported-harness placeholder whose directory is never read (2.629953ms)
-✔ unknown conversations, empty files and non-Pi files refuse (2.988026ms)
-✔ an unreadable file or root inside the roots is refused per row, not a failed catalogue (1.388415ms)
-✔ a seat directory without search permission refuses that root, not the catalogue (2.248255ms)
-✔ every page and cursor is a valid CHAT-01 record (847.042088ms)
-✔ the engine pin holds for the installed package (2.933293ms)
-✔ pinned Pi, sealed and without credentials, answers the controller's commands with the shapes the fake models (364.994239ms)
-✔ pinned Pi appends thinking_level_change at start when the branch lacks one, so the leaf moves (K8 then fails closed) (307.845583ms)
-✔ N25: ordinary Interrupt reconciles; a non-empty queue_update in the window is O5 (80.815781ms)
-✔ N1: an extension's follow-up queued after the fence is cleared before any abort; O5, Unknown (57.148386ms)
-✔ N1: a follow-up queued before the fence is O5 at once; the Interrupt refuses fenced (24.835014ms)
-✔ N2: with abort first, the fake runs the external item (the ordering guard has teeth) (21.576912ms)
-✔ N3: the fence lands in preflight, preflight errors, no run: failed, No run, uncertain (42.350718ms)
-✔ N4: the ack arrives after the first abort and a run starts: clear and abort again; Interrupted (37.456929ms)
-✔ N5: an input handler takes the prompt: ack, no run, delivery-unknown handled-without-run (119.71406ms)
-✔ N6: an extension queues between clear_queue and abort: O5 and O6, Unknown (46.194593ms)
-✔ N7: clear_queue times out: no abort, nativeQueue unknown, force stop still ends it (1530.004862ms)
-✔ N7: clear_queue answers an error: no abort, nativeQueue unknown, the link not poisoned (15.860593ms)
-✔ N8: an extension prompt starts a run during Mosaic preflight; the losing settle is O3 (71.692215ms)
-✔ N9: a run that started before the fence and ends aborted: failed interrupted, Interrupted (16.241237ms)
-✔ N9: decision 34: a run that ends aborted with no stop in progress: aborted-without-stop, uncertain, outcome unknown (16.048665ms)
-✔ N9: an aborted that lands after the fence but before any abort is written: aborted-without-stop, Unknown (32.548668ms)
-✔ N10: fake conformance (30.957332ms)
-✔ N11: the run fails before any user message_start: delivery-unknown ack-without-start, never failed (29.834953ms)
-✔ N12: input that starts a run after the final empty clear is O1 and not part of the stop's proof (19.905915ms)
-✔ N13: agent_start with no slot held is O1; a later prompt refuses with zero engine bytes (66.348845ms)
-✔ N14: the run completes while clear_queue is in flight: finished, Completed first, uncertain (26.705932ms)
-✔ N14: the run completes after the abort is written, before Pi applies it: finished, never relabelled (25.622164ms)
-✔ N15: the fence lands in preflight, then an input handler takes it: handled-without-run, No run (19.804122ms)
-✔ N16: Interrupt with no slot and no run refuses no-turn: no stop, no bytes, admission open (13.437887ms)
-✔ N17: the run fails on its own during the exchange: failed, Failed on its own (28.095743ms)
-✔ N18: no final assistant message_end, or a lost line: working stays working; before working, transport-unknown (113.466843ms)
-✔ N19: a losing extension prompt settles inside the Mosaic run before its user message: O3, run-overlap (134.279618ms)
-✔ N20: an extension triggerTurn during Mosaic preflight starts first; while streaming it queues with no signal (83.211293ms)
-✔ N21: a losing settle after the receipt settled finished is O2; the receipt stays finished (14.999139ms)
-✔ N22: an agent-level custom message is dropped by the clear with no signal; evidence names the seal (13.807672ms)
-✔ N23: a nextTurn message survives clear and abort and attaches to the next prompt, with no signal (13.596018ms)
-✔ N24: the seal is an allow-list: --extension, a missing --no-* flag, a second --mode or --session, a session or output flag, or a stray word refuses unsealed-engine; no engine starts (37.274093ms)
-ℹ tests 152
-ℹ suites 0
-ℹ pass 152
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 32488.500969

agents/darkwing/work/cohort-k1/runs/node-webui.txt

@@ -1,23 +0,0 @@
-✔ browser edge states: loading, empty, malformed, stale, hostile/long values, in-flight reply and appearance fallback (2364.731167ms)
-Rendered contrast: {"failures":[],"count":330,"lowest":4.504658476260286}
-✔ served Console browser: real board fixtures, keyboard, drafts, receipts, themes, 320px and failures (2725.60223ms)
-✔ conversation view: full history, collapsed tools, hidden thinking, inert hostile content, malformed and reconcile markers (2335.250549ms)
-✔ conversation view: a fork keeps the open branch, says so, and opens the new one on request (1375.477504ms)
-✔ conversation view: a newer session with no readable history keeps the marker (1002.908949ms)
-✔ conversation view: seats without history say so and offer no reply (608.42244ms)
-✔ Discord row through real board/WebUI: independent brake/liveness, no Reply, literal content (1957.413637ms)
-✔ return flow through the conversation view: send, tool call, delayed result, peer message, exact long answers, relaunch (53771.943793ms)
-✔ both presentations replace old activity with relaunch notice, label retained history, then resume after new activity (1987.210959ms)
-✔ reported return flow and relative Age: reply sent from the inspector, then the new answer appears there without manual refresh (21912.250845ms)
-✔ loopback host and board origin fail closed (7.055591ms)
-✔ real board fixture passes through WebUI; assets and isolated seen/reply work (100.024626ms)
-✔ proxy preserves exact request bytes, status and receipt, rejects forms and malformed JSON, never follows redirect (227.262245ms)
-✔ unreachable board reports URL; CLI rejects unsupported options (1632.267153ms)
-ℹ tests 14
-ℹ suites 0
-ℹ pass 14
-ℹ fail 0
-ℹ cancelled 0
-ℹ skipped 0
-ℹ todo 0
-ℹ duration_ms 54170.048795

agents/darkwing/work/cohort-k1/runs/start.txt

@@ -1,2 +0,0 @@
-2026-10-09T13:05:51Z
- 08:05:51 up 33 days,  9:40,  5 users,  load average: 4.03, 7.84, 5.58

agents/darkwing/work/cohort-k1/scope-receipt.mjs

@@ -1,61 +0,0 @@
-// Row 46 receipt probe. Runs the fixture's ignoreTerm child inside a
-// delegated systemd user scope, sends it SIGTERM, and records the scope's
-// `systemctl --user show`, `cgroup.procs` before and after TERM, and the
-// child's exit code and signal.
-//
-//   node scope-receipt.mjs <mode> <delayMs> <runs>
-//
-// mode `race`: TERM goes `delayMs` after spawn, as the fixture does today.
-// mode `ready`: TERM goes `delayMs` after the child reports its handler is
-// installed, as the fixed fixture does.
-// The outer process launches one scope per run; the inner process
-// (`--inner`) is the scope's main process and prints one JSON line.
-
-import { spawn, spawnSync } from "node:child_process";
-import { readFileSync } from "node:fs";
-
-const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
-
-// The fixture's child code for { ignoreTerm: true } with no pidLog
-// (packages/conversation/tests/fake-pi.mjs, spawnChild), plus a readiness
-// byte on stdout in `ready` mode.
-const childCode = (ready) =>
-  "const note=()=>{};process.on('SIGTERM',()=>note('term'));" +
-  (ready ? "process.stdout.write('r');" : "") +
-  "setInterval(()=>{},1e9)";
-
-const procs = (cg) => readFileSync(`/sys/fs/cgroup${cg}/cgroup.procs`, "utf8").split("\n").filter(Boolean).map(Number);
-
-async function inner(mode, delayMs, unit) {
-  const cg = readFileSync("/proc/self/cgroup", "utf8").trim().split("::")[1];
-  const ready = mode === "ready";
-  const t0 = performance.now();
-  const child = spawn(process.execPath, ["-e", childCode(ready)], { stdio: ["ignore", ready ? "pipe" : "ignore", "ignore"] });
-  const exited = new Promise((r) => child.on("exit", (code, signal) => r({ code, signal, atMs: Math.round(performance.now() - t0) })));
-  let readyMs = null;
-  if (ready) {
-    await new Promise((r) => child.stdout.once("data", r));
-    readyMs = Math.round(performance.now() - t0);
-  }
-  await sleep(delayMs);
-  const show = spawnSync("systemctl", ["--user", "show", "-p", "Id,LoadState,ActiveState,SubState,InvocationID,ControlGroup,Delegate", `${unit}.scope`], { encoding: "utf8" }).stdout.trim().split("\n");
-  const before = procs(cg);
-  const termMs = Math.round(performance.now() - t0);
-  child.kill("SIGTERM");
-  const exit = await Promise.race([exited, sleep(500).then(() => null)]);
-  const after = procs(cg);
-  if (!exit) child.kill("SIGKILL");
-  console.log(JSON.stringify({ mode, delayMs, readyMs, termMs, child: child.pid, self: process.pid, show, before, after, childInBefore: before.includes(child.pid), childInAfter: after.includes(child.pid), exit: exit ?? "alive 500 ms after TERM" }));
-}
-
-async function outer(mode, delayMs, runs) {
-  for (let i = 0; i < runs; i++) {
-    const unit = `dw-r46-${mode}-${delayMs}-${process.pid}-${i}`;
-    const r = spawnSync("systemd-run", ["--user", "--scope", "--quiet", "-p", "Delegate=yes", `--unit=${unit}`, "--", process.execPath, import.meta.filename, "--inner", mode, String(delayMs), unit], { encoding: "utf8" });
-    process.stdout.write(r.stdout || `{"unit":"${unit}","status":${r.status},"stderr":${JSON.stringify(r.stderr)}}\n`);
-  }
-}
-
-const a = process.argv.slice(2);
-if (a[0] === "--inner") await inner(a[1], Number(a[2]), a[3]);
-else await outer(a[0], Number(a[1]), Number(a[2] ?? 5));

agents/darkwing/work/cohort-k1/trace-fail.jsonl

@@ -1,9 +0,0 @@
-{"t":1791551027129,"ev":"spawn","pid":3646877,"ignoreTerm":true,"setsid":true}
-{"t":1791551027142,"ev":"term","pid":3646862}
-{"t":1791551027142,"ev":"term","pid":3646877}
-{"t":1791551029516,"ev":"spawn","pid":3647263,"ignoreTerm":true,"setsid":false}
-{"t":1791551029528,"ev":"term","pid":3647248}
-{"t":1791551029528,"ev":"term","pid":3647263}
-{"t":1791551031873,"ev":"spawn","pid":3647581,"ignoreTerm":true,"setsid":false}
-{"t":1791551031886,"ev":"term","pid":3647556}
-{"t":1791551031887,"ev":"term","pid":3647581}

agents/darkwing/work/cohort-k1/trace-load.jsonl

@@ -1,13 +0,0 @@
-{"t":1791550812775,"ev":"spawn","pid":3607494,"ignoreTerm":true,"setsid":true}
-{"t":1791550812853,"ev":"ready","pid":3607494}
-{"t":1791550812878,"ev":"term","pid":3607446}
-{"t":1791550812878,"ev":"term","pid":3607494}
-{"t":1791550815693,"ev":"spawn","pid":3607807,"ignoreTerm":true,"setsid":false}
-{"t":1791550815773,"ev":"ready","pid":3607807}
-{"t":1791550815779,"ev":"term","pid":3607781}
-{"t":1791550815780,"ev":"term","pid":3607807}
-{"t":1791550819446,"ev":"spawn","pid":3608344,"ignoreTerm":true,"setsid":false}
-{"t":1791550819489,"ev":"ready","pid":3608344}
-{"t":1791550819556,"ev":"term","pid":3608301}
-{"t":1791550819556,"ev":"term","pid":3608344}
-{"t":1791550820058,"ev":"term","pid":3608344}

agents/darkwing/work/cohort-k1/trace-pass.jsonl

@@ -1,13 +0,0 @@
-{"t":1791550802745,"ev":"spawn","pid":3605642,"ignoreTerm":true,"setsid":true}
-{"t":1791550802763,"ev":"ready","pid":3605642}
-{"t":1791550802798,"ev":"term","pid":3605597}
-{"t":1791550802798,"ev":"term","pid":3605642}
-{"t":1791550805434,"ev":"spawn","pid":3606162,"ignoreTerm":true,"setsid":false}
-{"t":1791550805456,"ev":"ready","pid":3606162}
-{"t":1791550805492,"ev":"term","pid":3606153}
-{"t":1791550805492,"ev":"term","pid":3606162}
-{"t":1791550808090,"ev":"spawn","pid":3606502,"ignoreTerm":true,"setsid":false}
-{"t":1791550808106,"ev":"ready","pid":3606502}
-{"t":1791550808199,"ev":"term","pid":3606487}
-{"t":1791550808200,"ev":"term","pid":3606502}
-{"t":1791550808539,"ev":"term","pid":3606502}

agents/darkwing/work/cohort-k1/trace.patch

@@ -1,43 +0,0 @@
-diff --git a/packages/conversation/src/shim.mjs b/packages/conversation/src/shim.mjs
-index 2adbe48b..d51c4a1f 100644
---- a/packages/conversation/src/shim.mjs
-+++ b/packages/conversation/src/shim.mjs
-@@ -22,7 +22,7 @@
- // `populated 0`. A missing or unreadable file is unavailable, never empty.
- 
- import { spawn } from "node:child_process";
--import { closeSync, mkdirSync, readdirSync, readFileSync, unlinkSync, writeFileSync } from "node:fs";
-+import { appendFileSync, closeSync, mkdirSync, readdirSync, readFileSync, unlinkSync, writeFileSync } from "node:fs";
- import { createServer } from "node:net";
- import { join } from "node:path";
- import { LineSplitter, encodeLine, parseLine } from "./framing.mjs";
-@@ -136,6 +136,7 @@ async function handle(req) {
-         if (startOf(pid) !== startTicks) continue;
-         try {
-           process.kill(pid, "SIGTERM");
-+          if (process.env.DW_TRACE) appendFileSync(process.env.DW_TRACE, JSON.stringify({ t: Date.now(), ev: 'term', pid }) + '\n');
-           signalled.push(pid);
-         } catch {
-           // gone already
-diff --git a/packages/conversation/tests/fake-pi.mjs b/packages/conversation/tests/fake-pi.mjs
-index b3014f86..f30f5a4c 100644
---- a/packages/conversation/tests/fake-pi.mjs
-+++ b/packages/conversation/tests/fake-pi.mjs
-@@ -615,13 +615,16 @@ const children = [];
- // K2); `forkLoop` forks every 5 ms (K12); `ignoreTerm` survives SIGTERM, so
- // only the kill phase ends it (K3, K10, K11). With `pidLog`, the fork loop
- // appends each child's pid and a `term` line when it gets SIGTERM (K12).
-+import * as __fs from 'node:fs';
-+const require0 = () => __fs;
- function spawnChild({ setsid = false, forkLoop = false, ignoreTerm = false, pidLog = null } = {}) {
-   const note = pidLog ? `const note=(s)=>require('node:fs').appendFileSync(${JSON.stringify(pidLog)},s+'\\n');` : "const note=()=>{};";
--  const code = note + (ignoreTerm ? "process.on('SIGTERM',()=>note('term'));" : "") + (forkLoop
-+  const code = note + (ignoreTerm ? "process.on('SIGTERM',()=>note('term'));" + (process.env.DW_TRACE ? `require('node:fs').appendFileSync(${JSON.stringify(process.env.DW_TRACE)},JSON.stringify({t:Date.now(),ev:'ready',pid:process.pid})+'\\n');` : "") : "") + (forkLoop
-     ? "const {spawn}=require('node:child_process');setInterval(()=>{try{const c=spawn('sleep',['1000'],{stdio:'ignore'});if(c.pid)note(String(c.pid))}catch{}},5);setInterval(()=>{},1e9)"
-     : "setInterval(()=>{},1e9)");
-   const child = spawn(process.execPath, ["-e", code], { stdio: "ignore", detached: setsid });
-   children.push(child.pid);
-+  if (process.env.DW_TRACE) { require0().appendFileSync(process.env.DW_TRACE, JSON.stringify({ t: Date.now(), ev: 'spawn', pid: child.pid, ignoreTerm, setsid }) + '\n'); child.on('exit', (code, sig) => require0().appendFileSync(process.env.DW_TRACE, JSON.stringify({ t: Date.now(), ev: 'child-exit', pid: child.pid, code, sig }) + '\n')); }
-   return child.pid;
- }
- 

agents/darkwing/work/discord-board/acceptance.md

@@ -1,28 +0,0 @@
-# Independent acceptance checklist, row 18
-
-Darkwing reviews Filbert's implementation without editing its source candidate.
-Dewey reviews visible connector presentation. No live connector manipulation.
-
-- Discovery accepts only safe matching binding name/seat from private regular
-  files, never dereferences a token path and never serializes private fields.
-- Path traversal, symlinked binding/runtime/session paths and malformed records
-  cannot cause arbitrary reads or an actionable/live row.
-- No owner, malformed owner, dead PID, missing identity, reused PID and boot
-  mismatch are non-live. A positively matching live process is live.
-- STOP presence is visible as braked independently of process liveness. Its
-  contents are not read or exposed; no STOP or lock is created or changed.
-- Ordinary completed messages remain idle. No false human attention regression.
-- Connector rows cannot borrow a native agent's registration for replies.
-  Exercise replyToRow and HTTP using a fake executable hook; every connector
-  attempt must be refused before that hook runs, including with forged tmux
-  registration. Normal-agent reply tests must still pass.
-- Both existing board and WebUI distinguish the connector and brake state and
-  omit reply controls. Preserve escaping, including hostile binding fixtures.
-- Discovery errors disclose no private JSON fields or raw contents. One bad
-  binding must not silently manufacture a healthy row.
-- Candidate pins match before and after tests. Existing dirty attention changes
-  remain intact; no unrelated source integration or live operation is inferred.
-
-After source approval, measure the real row read-only. Offline/braked behavior
-uses isolated fixtures unless the operator separately approves a live-service
-transition. Board replacement is its own protected gate.

agents/darkwing/work/discord-board/backend-live-verification.json

@@ -1,23 +0,0 @@
-{
-  "at": "2026-09-14T13:51:10.530662+00:00",
-  "backendPid": 3769124,
-  "health": "ok",
-  "row": {
-    "agent": "sage (discord: shared-signals)",
-    "project": "fleet",
-    "state": "idle",
-    "alive": true,
-    "connector": {
-      "binding": "shared-signals",
-      "braked": false,
-      "ownerState": "live",
-      "alive": true
-    },
-    "task": "Discord connector",
-    "taskSource": "connector"
-  },
-  "replyStatus": 409,
-  "replyError": "board replies are disabled for Discord connectors",
-  "fiveAgentPaneIdentitiesUnchanged": true,
-  "connectorServiceIdentityUnchanged": true
-}

agents/darkwing/work/discord-board/backend-restart.jsonl

@@ -1,2 +0,0 @@
-{"at": "2026-09-14T13:50:24.078636+00:00", "event": "owner-authorized-restart-intent", "oldPid": 3204655, "agents": {"default/darkwing": [["2733924", "12863634"]], "default/dewey": [["934346", "466065"]], "default/filbert": [["72183", "100870"]], "default/researcher": [["173699", "66404285"]], "mosaic-fleet/rocko": [["90599", "128275"]]}, "connectorService": [3022843, "67887873"], "manifest": "254403b89c0a2330da53e8dbad1cbeba3b1b06cf4f3efddc18451e04cb78f6de"}
-{"at": "2026-09-14T13:50:24.503231+00:00", "event": "replacement-started", "oldExitedGracefully": true, "newPid": 3769124, "log": "/tmp/discord-board-backend-ovk_cahk.log"}

agents/darkwing/work/discord-board/r1-backend-review.json

@@ -1,18 +0,0 @@
-{
-  "at": "2026-09-14T01:10:19.375709+00:00",
-  "candidate": "/tmp/discord-board-r1-KbMrGQWF",
-  "manifestSha256": "5c92acc90d202727d790f3fb8d74387db1c9e5c3e43d4f4b56b40e5ae503a56a",
-  "verdict": "CHANGES REQUIRED",
-  "independentSerializedTests": 320,
-  "finding": {
-    "id": "R1-B1",
-    "severity": "P2",
-    "file": "packages/control-board/src/discord.mjs",
-    "issue": "STOP metadata access errors collapse to absence, falsely projecting not braked",
-    "reproduction": "Synthetic journal directory contains STOP, chmod directory to 000 as uid 1000, inspectDiscord returns braked:false, ownerState:invalid, alive:false. Restore permissions and remove fixture.",
-    "expected": "braked:null/unknown when STOP existence cannot be established; false only for verified absence",
-    "required": "Distinguish missing metadata from access errors and add non-root unreadable-directory regression."
-  },
-  "ux": "Dewey APPROVE on exact R1; three independent serialized browser tests passed, source/automation limitations retained",
-  "parallelQualification": "Two author concurrent frozen timeouts remain unresolved and are not green; serialized independent run passed."
-}

agents/darkwing/work/discord-board/r2-envelope-review.json

@@ -1,7 +0,0 @@
-{
-  "candidate": "R2",
-  "syntheticOnly": true,
-  "taskContainsEnvelopeAuthorId": true,
-  "taskContainsEnvelopeMessageId": true,
-  "taskSource": "first-user-message"
-}

agents/darkwing/work/discord-board/r3-backend-review.json

@@ -1,18 +0,0 @@
-{
-  "at": "2026-09-14T01:26:42.688410+00:00",
-  "candidate": "/tmp/discord-board-r3-U9vVrlQu",
-  "manifestSha256": "254403b89c0a2330da53e8dbad1cbeba3b1b06cf4f3efddc18451e04cb78f6de",
-  "reviewer": "Darkwing",
-  "backendVerdict": "APPROVE AS SOURCE",
-  "verified": "Nine working/frozen pins, exact three-file R2-to-R3 delta, inherited attention pins and full serialized six-package suite 322/322",
-  "findingsClosed": [
-    "R1-B1: inaccessible STOP is unknown, non-root regression passes",
-    "R2-B2: canonical routing envelope no longer becomes connector Task; ordinary fallback retained"
-  ],
-  "limitations": [
-    "No generalized transcript redaction",
-    "R1 concurrent combined frozen timeouts unresolved/not green",
-    "No live observation, backend restart, connector change or publication in this review"
-  ],
-  "uxGate": "Await exact R3 confirmation from Dewey via agent-send"
-}

agents/darkwing/work/discord-board/r3-live-observation.json

@@ -1,20 +0,0 @@
-{
-  "at": "2026-09-14T01:28:17.695Z",
-  "sourceApproval": 26257,
-  "readOnly": true,
-  "agent": "sage (discord: shared-signals)",
-  "project": "fleet",
-  "state": "idle",
-  "alive": true,
-  "connector": {
-    "binding": "shared-signals",
-    "braked": false,
-    "ownerState": "live",
-    "alive": true
-  },
-  "task": "Discord connector",
-  "taskSource": "connector",
-  "registrationAbsent": true,
-  "ownerMatchesService": true,
-  "discoveryErrorCount": 0
-}

agents/darkwing/work/discord-board/reply-baseline.json

@@ -1,9 +0,0 @@
-{
-  "at": "2026-09-14T00:50:35.339Z",
-  "sourceSha256": "dfbb7ab9374c0ac9fafa0503f495abd938f499f5d6227233de03a60ea3022927",
-  "fixture": "connector row with forged native registration",
-  "status": 200,
-  "fakeTransportCalls": 1,
-  "realTransportCalls": 0,
-  "gatePassed": false
-}

agents/darkwing/work/discord-engine-busy/README.md

@@ -1,162 +0,0 @@
-# Discord engine: guaranteed test cleanup and the timeout gap in `busy` (#1509), R2 candidate
-
-Sage assigned this on 2026-09-26 as 6b, source only. Rocko reviews. Base is
-HEAD 401cc850. Not committed. The live connector runs from this checkout, so
-Sage is holding its restart until this is approved and committed. Nobody
-should restart it from a working copy.
-
-## Defects (DEFERRED Open, "Discord engine: leaked fake pi…")
-
-(a) `engine.test.mjs` read the fake's `commands.jsonl` 20 ms after a prompt and
-got ENOENT under load. Seven tests stopped the engine outside `finally`, so a
-failed assertion left the fake pi running and the test file never exited.
-
-(b) `busy` was `state.busy || pending.some((t) => !t.done)`. If a turn timed out
-before its `agent_start` was read, it was done while `state.busy` was still
-false. The next prompt then went straight to pi, which refused it as
-streaming.
-
-## R1 and Rocko's finding
-
-R1 held later prompts behind a failed turn. If pi had sent no `agent_start` for
-it within a grace period, R1 dropped that turn from the queue and sent the next
-prompt. Rocko rejected it (F1, High), in
-`agents/rocko/work/discord-engine-busy-r1-review-2026-09-26.md`, sha256
-047dbd8f.
-
-Pi's events carry no prompt id. The engine attributes them to the front of its
-queue. Silence until the grace ends does not prove the old run will never come.
-If pi then runs it, its events land on the new prompt, which R1 had just put at
-the front. Rocko's reproducer got the old run's answer and its `old.md` tool
-record back as the new prompt's result. My R1 README said such events "find no
-live head and are dropped". That was wrong.
-
-The R1 files stay here as `r1-manifest.sha256` and `r1.patch`.
-
-## Change (R2)
-
-`packages/discord/src/engine-pi.mjs`:
-- `engineBusy()` is `state.busy || state.pending.length > 0`. A failed turn
-  still in the queue holds the next prompt back, and stays at the front, so any
-  late events for it land on it. `prompt()`, `sendHeld()` and the `busy` getter
-  use it. This part is unchanged from R1.
-- The bound is now a stop, not a drop. When a turn fails while it is still in
-  the queue, `failTurn` starts a timer, `abortGraceMs` (default
-  `ABORT_GRACE_MS`, 30 s, an engine option, not binding config). When it
-  fires:
-  - If pi has sent `agent_start` (`state.busy`), nothing happens. That run
-    ends on its `agent_end` or a settle, as on HEAD.
-  - Otherwise `wedge()` sets `state.wedged`, fails every held prompt with
-    code `engine-wedged`, and stops pi: stdin closed, SIGTERM, then SIGKILL
-    after 5 s. The failed turn stays at the front until the exit.
-- While wedged, nothing is written to that child. `write()`, `sendHeld()` and
-  `prompt()` refuse, and a new prompt fails at once with `engine-down`. The exit
-  runs the usual `failAll` and `onExit`.
-- `stop()`'s body moved into `stopChild()`, which both `stop()` and `wedge()`
-  call.
-- `release()` clears the timer wherever a turn leaves the queue: `agent_end`,
-  settle, a refused send, and process exit. As in R1, the settle handler removes
-  turns before failing them.
-
-What recovery looks like live: `cli.mjs` handles `onExit` with `shutdown(1)`.
-The unit's `Restart=on-failure` starts a new connector and a new pi 15 s later,
-within its limit of five tries in ten minutes. This change doesn't touch the
-unit or the restart policy. A wedge now costs one connector restart. R1 would
-have kept the same pi and risked a wrong answer.
-
-`packages/discord/tests/fake-pi.mjs`:
-- `mute`: accepted and never run.
-- `stall <ms>`: accepted, then the fake reads nothing for `<ms>`, runs the
-  stalled prompt, and only then reads what came in meanwhile. This is the
-  order in Rocko's case.
-
-`packages/discord/tests/engine.test.mjs`:
-- `withEngine()` stops the engine in `finally`. Every test that starts an
-  engine uses it, or has its own `try/finally` in the exit test.
-- `commands()` returns `[]` until the fake creates its log. The held-prompt
-  test waits for the first prompt with `until()` instead of a 20 ms sleep, and
-  its first prompt is `slow 300`.
-- The manual-timer test fires the turn timer before any pi event is read. The
-  next prompt must wait for the settle and get its own answer.
-- New or changed for R2:
-  - `mute` with `abortGraceMs: 150`. The held prompt fails with
-    `engine-wedged` after the grace, a later prompt fails with
-    `engine-down`, `onExit` fires, and pi saw only `mute` and `abort`.
-  - `stall 400` with the same grace, which is Rocko's case with a real
-    child. The held prompt fails with `engine-wedged`, pi exits, and "after
-    stall" never reaches pi.
-  - Rocko's reproducer as an in-memory test, run twice. The old prompt's
-    response comes either before its timeout or only with the late events.
-    After the grace, the old run's start, tool pair, answer, end and settle
-    arrive while pi is still exiting. The held prompt stays failed with
-    `engine-wedged` and a later prompt fails with `engine-down`. Pi saw only
-    `old` and `abort`, then SIGTERM, then SIGKILL at 5 s. Only the exit
-    reaches `onExit`. The test reads recorded outcomes after a tick instead
-    of awaiting, so a regression fails instead of hanging.
-  - `late 400` with the same grace. Pi started that run, so the grace does
-    not stop pi, and the next prompt gets its own answer when the run ends.
-
-## Evidence
-
-- `engine.test.mjs`: 17/17.
-- R1's engine (d5bf24b5, from `r1.patch`) against these tests fails 4: `mute`,
-  `stall`, and both in-memory runs. In that run a probe shows R1 answering
-  "after stall" with "echo: stalled". An earlier draft of the in-memory test
-  awaited the held prompt and hung on R1 until the 120 s cap. It now fails in
-  milliseconds.
-- HEAD's engine against these tests fails 5: the manual-timer test and the
-  same four.
-- Mutations of R2:
-  - Without the `state.busy` check, the `late 400` test fails.
-  - Without the `wedge()` call, 4 fail.
-  - Without failing held prompts in `wedge()`, 4 fail.
-- Test union (control-board, webui, seat, mosaic, ledger, discord) at default
-  concurrency on `git archive` of 401cc850 plus the three files: 406/406
-  three times, 23 to 24 s each. No fake pi was left running.
-- Eight suites green on that snapshot: config 24, task 90, foundation 43,
-  conductor 17, release 14, auth 15, discord 63, extension-package 18.
-
-Logs: `/tmp/dw-6b-r2-conc-{1,2,3}.txt`. R1's evidence runs:
-`/tmp/dw-6b-conc-{1,2,3}.txt`, `/tmp/dw-6b-serial.txt`. HEAD's hang control:
-`/tmp/dw-1509-headctl-{1,2,3}.txt`, `/tmp/dw-1509-ef00-1.txt`. There, HEAD hit
-the 240 s cap at 199 ok under the union's load.
-
-## Not covered
-
-- A run pi started and never ends, even after abort, still holds prompts
-  until pi settles or exits. Each held prompt fails at its own timeout ("while
-  waiting for the engine"). HEAD behaves the same way through `state.busy`, and
-  Rocko did not block on it. Only a pi restart clears it.
-- A wedge ends the connector process, and the recovery is systemd's restart.
-  Nothing here changes the unit, and the restart limit still applies.
-- No live restart, and no change to the binding schema.
-
-## Frozen files
-
-`r2-manifest.sha256` holds the three R2 hashes. `r2.patch` is `git diff
-packages/discord` at freeze time.
-
-## Review
-
-Rocko, R1, 2026-09-26: request changes, F1 High, as described above. Report:
-`agents/rocko/work/discord-engine-busy-r1-review-2026-09-26.md`, sha256
-047dbd8f.
-
-Rocko, R2, 2026-09-26: approved the three pinned files. Report:
-`agents/rocko/work/discord-engine-busy-r2-review-2026-09-26.md`, sha256
-ed5510a0. He checked the manifests before and after, ran 17/17 himself, and
-read the CLI shutdown path, `connector.stop` and the unit template. Sage asked
-him three operational questions:
-- A wedge exits 1, never 3. Exit 3 remains the supervised startup refusal.
-- The unit's start limit (5 starts in 600 s) is a rate limit. It does not
-  bound repeated wedges. With the default 180 s turn timeout, the 30 s grace
-  and the 15 s restart delay, a cycle takes at least 225 s. That stays under
-  the limit, so a pi that wedges every time could restart indefinitely.
-  Stopping for good after repeated wedges would need a separate policy. This
-  change does not add one.
-- He recommends, as a nonblocking follow-up, that the connector journal
-  record at startup: HEAD, dirty state scoped to runtime source, and a digest
-  of the runtime files. A wedge restart loads whatever the checkout holds.
-
-This section was added after approval, so the README hash no longer matches
-the one Rocko pinned (69350f29). The three source files are unchanged.

agents/darkwing/work/discord-engine-busy/r1-manifest.sha256

@@ -1,3 +0,0 @@
-d5bf24b59c07c85067f4087c03b54ca8b4df923c1d591441dedd2e8a7ff2ae39  packages/discord/src/engine-pi.mjs
-f0abee9c243d46d66dd2271abc3fd89089c350ac6a66ab49131bce80adfcdc33  packages/discord/tests/engine.test.mjs
-fa1bf44e3f33eb970a714ada1c686abbc1932baaf418679276edf8813abbe6de  packages/discord/tests/fake-pi.mjs

agents/darkwing/work/discord-engine-busy/r1.patch

@@ -1,414 +0,0 @@
-diff --git a/packages/discord/src/engine-pi.mjs b/packages/discord/src/engine-pi.mjs
-index 5c8fd0a9..9dcaeb42 100644
---- a/packages/discord/src/engine-pi.mjs
-+++ b/packages/discord/src/engine-pi.mjs
-@@ -16,9 +16,12 @@
- // from `tool_execution_start`/`tool_execution_end` into the result so the
- // turn record shows what was read. An `agent_end` with `willRetry` is not
- // the end of the run. A timeout sends `abort` and fails that turn; the
--// process stays. A malformed JSONL line from pi fails the current turn (its
--// outcome is now unknowable) and the process stays. Process exit fails
--// every pending turn and is reported through `onExit`.
-+// process stays. The failed turn holds later prompts back until its
-+// agent_end or a settle. If pi has not started it within ABORT_GRACE_MS, it
-+// is dropped and the next prompt goes out; a run pi did start holds them
-+// until it ends, as any run does. A malformed JSONL line from pi fails the
-+// current turn (its outcome is now unknowable) and the process stays.
-+// Process exit fails every pending turn and is reported through `onExit`.
- //
- // Framing follows pi's RPC doc: split on "\n" only, strip a trailing "\r".
- // Node readline is not used because it also splits on U+2028/U+2029.
-@@ -64,12 +67,19 @@ export function assistantText(message) {
-     .trim();
- }
- 
-+// How long a turn that failed here (timeout, protocol error) may wait for
-+// pi's agent_start before it stops holding the next prompt back. Without a
-+// bound, a prompt pi accepted but never ran would queue every later prompt
-+// until restart.
-+export const ABORT_GRACE_MS = 30000;
-+
- export function createEngine({
-   command, args, cwd, env = {},
-   spawn = nodeSpawn,
-   setTimeoutImpl = globalThis.setTimeout, clearTimeoutImpl = globalThis.clearTimeout,
-   log = () => {},
-   onExit = () => {},
-+  abortGraceMs = ABORT_GRACE_MS,
- } = {}) {
-   if (typeof command !== "string" || command.length === 0) throw new DiscordError("engine: command required", 1);
-   if (!Array.isArray(args)) throw new DiscordError("engine: args required", 1);
-@@ -80,15 +90,35 @@ export function createEngine({
- 
-   // A turn that fails on the client side (timeout, protocol error) stays in
-   // the pending queue, marked done, until pi's own turn_end for it arrives.
--  // Otherwise that turn_end would be attributed to the next prompt.
-+  // Otherwise that turn_end would be attributed to the next prompt. It holds
-+  // later prompts back; if pi has not started it within abortGraceMs, it goes.
-   function failTurn(turn, code, message) {
-     if (turn.done) return;
-     turn.done = true;
-     if (turn.timer !== null) clearTimeoutImpl(turn.timer);
-     turn.timer = null;
-+    if (state.pending.includes(turn)) {
-+      turn.grace = setTimeoutImpl(() => {
-+        turn.grace = null;
-+        // No agent_start by now: pi never started this run and will send no
-+        // agent_end for it, so it leaves the queue and cannot take the next
-+        // prompt's. A run pi did start keeps its place until it ends.
-+        if (!state.busy) {
-+          const i = state.pending.indexOf(turn);
-+          if (i !== -1) state.pending.splice(i, 1);
-+        }
-+        sendHeld();
-+      }, abortGraceMs);
-+    }
-     turn.reject(new DiscordError(message, 1, { code }));
-   }
- 
-+  // Call when a turn leaves the pending queue.
-+  function release(turn) {
-+    if (turn.grace !== null) clearTimeoutImpl(turn.grace);
-+    turn.grace = null;
-+  }
-+
-   function settleTurn(turn, value) {
-     if (turn.done) return;
-     turn.done = true;
-@@ -99,7 +129,10 @@ export function createEngine({
- 
-   function failAll(code, message) {
-     const pending = state.pending.splice(0);
--    for (const t of pending) failTurn(t, code, message);
-+    for (const t of pending) {
-+      release(t);
-+      failTurn(t, code, message);
-+    }
-     for (const h of state.held.splice(0)) failTurn(h.turn, code, message);
-     for (const [, r] of state.responses) r.reject(new DiscordError(message, 1, { code }));
-     state.responses.clear();
-@@ -174,6 +207,7 @@ export function createEngine({
-       // Attribute the run to the head even if it failed client-side, so the
-       // next prompt's agent_end is not taken for this one.
-       const run = state.pending.shift();
-+      if (run) release(run);
-       if (!run || run.done) return;
-       const messages = Array.isArray(event.messages) ? event.messages.filter((m) => m && m.role === "assistant") : [];
-       const message = messages.length > 0 ? messages[messages.length - 1] : run.last;
-@@ -193,13 +227,14 @@ export function createEngine({
-       // this settle and still has no agent_end will never get one: fail it now
-       // instead of waiting for its timeout. Turns whose prompt response has
-       // not arrived yet belong to a later run and stay.
-+      const dropped = [];
-       const keep = [];
--      for (const t of state.pending) {
--        if (t.done) continue;
--        if (t.accepted) failTurn(t, "engine-settled-without-turn", "engine settled without answering this prompt");
--        else keep.push(t);
--      }
-+      for (const t of state.pending) (t.done || t.accepted ? dropped : keep).push(t);
-       state.pending = keep;
-+      for (const t of dropped) {
-+        release(t);
-+        failTurn(t, "engine-settled-without-turn", "engine settled without answering this prompt");
-+      }
-       sendHeld();
-     }
-   }
-@@ -214,15 +249,23 @@ export function createEngine({
-       // Never accepted: pi will not emit a turn_end for it, so remove it.
-       const i = state.pending.indexOf(turn);
-       if (i !== -1) state.pending.splice(i, 1);
-+      release(turn);
-       failTurn(turn, (err.details && err.details.code) || "engine-refused", err.message);
-       sendHeld();
-     });
-   }
- 
-+  // Pi is busy from our side while any sent prompt is still queued, even one
-+  // that already failed here: a turn that timed out before its agent_start
-+  // was read leaves state.busy false while pi runs it, and sending then would
-+  // be refused as streaming. It leaves the queue on its agent_end, on a
-+  // settle, on a refused send, or when its grace ends before pi started it.
-+  const engineBusy = () => state.busy || state.pending.length > 0;
-+
-   // After a settle (or a refused send) the oldest held prompt goes out.
-   function sendHeld() {
-     if (state.exited !== null) return;
--    if (state.busy || state.pending.some((t) => !t.done)) return;
-+    if (engineBusy()) return;
-     const next = state.held.shift();
-     if (next) send(next.turn, next.command);
-   }
-@@ -281,7 +324,7 @@ export function createEngine({
-     // with DiscordError carrying details.code for the turn record.
-     prompt(text, { timeoutMs = 180000 } = {}) {
-       if (typeof text !== "string" || text.length === 0) throw new DiscordError("prompt text required", 1);
--      const turn = { resolve: null, reject: null, timer: null, done: false, accepted: false, tools: new Map(), turns: 0, last: null };
-+      const turn = { resolve: null, reject: null, timer: null, grace: null, done: false, accepted: false, tools: new Map(), turns: 0, last: null };
-       const done = new Promise((resolve, reject) => {
-         turn.resolve = resolve;
-         turn.reject = reject;
-@@ -310,13 +353,13 @@ export function createEngine({
-         failTurn(turn, "engine-down", "engine is not running");
-         return done;
-       }
--      if (state.busy || state.pending.some((t) => !t.done) || state.held.length > 0) state.held.push({ turn, command });
-+      if (engineBusy() || state.held.length > 0) state.held.push({ turn, command });
-       else send(turn, command);
-       return done;
-     },
- 
-     get busy() {
--      return state.busy || state.pending.some((t) => !t.done) || state.held.length > 0;
-+      return engineBusy() || state.held.length > 0;
-     },
-     get pendingCount() {
-       return state.pending.filter((t) => !t.done).length + state.held.length;
-diff --git a/packages/discord/tests/engine.test.mjs b/packages/discord/tests/engine.test.mjs
-index 59674d1e..62dd6017 100644
---- a/packages/discord/tests/engine.test.mjs
-+++ b/packages/discord/tests/engine.test.mjs
-@@ -28,7 +28,20 @@ function start(root, extra = {}) {
-     log: (m) => logs.push(m), ...extra,
-   });
-   engine.start();
--  return { engine, logs, commands: () => readFileSync(logPath, "utf8").trim().split("\n").filter(Boolean).map((l) => JSON.parse(l)) };
-+  // The fake creates its log on the first command; until then there are none.
-+  const commands = () => (existsSync(logPath) ? readFileSync(logPath, "utf8").trim().split("\n").filter(Boolean).map((l) => JSON.parse(l)) : []);
-+  return { engine, logs, commands };
-+}
-+
-+// Every test stops its engine in finally: a fake pi left running after a
-+// failed assertion keeps the test file from exiting.
-+async function withEngine(extra, body) {
-+  const started = start(makeRoot(), extra);
-+  try {
-+    await body(started);
-+  } finally {
-+    await started.engine.stop();
-+  }
- }
- 
- test("engine: buildPiArgs carries the fixed flags, engine settings, session dir and prompt file", () => {
-@@ -56,8 +69,7 @@ test("engine: with tools, buildPiArgs turns pi's own tools off, loads the extens
-   assert.equal(rw[rw.indexOf("--tools") + 1], "list_dir,read_file,search,write_file,edit_file", "a writable root adds exactly the two write tools");
- });
- 
--test("engine: a run with tool turns settles once, on the answer, with every tool call in the result", async () => {
--  const { engine } = start(makeRoot());
-+test("engine: a run with tool turns settles once, on the answer, with every tool call in the result", () => withEngine({}, async ({ engine }) => {
-   const r = await engine.prompt("tools 3");
-   assert.equal(r.text, "read 3 file(s)");
-   assert.equal(r.turns, 2);
-@@ -71,45 +83,38 @@ test("engine: a run with tool turns settles once, on the answer, with every tool
-   assert.equal(plain.turns, 1);
-   await idle(engine);
-   assert.equal(engine.busy, false);
--  await engine.stop();
--});
-+}));
- 
--test("engine: a run that ends on a tool-only turn fails the prompt as empty; a retried run settles on the real end", async () => {
--  const { engine } = start(makeRoot());
-+test("engine: a run that ends on a tool-only turn fails the prompt as empty; a retried run settles on the real end", () => withEngine({}, async ({ engine }) => {
-   const r = await engine.prompt("toolonly");
-   assert.equal(r.text, "", "no text: the connector turns this into engine-empty");
-   assert.equal(r.tools.length, 1);
-   const again = await engine.prompt("retry");
-   assert.equal(again.text, "after retry");
--  await engine.stop();
--});
-+}));
- 
--test("engine: one prompt, one turn, text and usage come back", async () => {
--  const { engine } = start(makeRoot());
--  try {
--    const r = await engine.prompt("hello");
--    assert.equal(r.text, "echo: hello");
--    assert.deepEqual(r.usage, { input: 3, output: 2 });
--    await idle(engine);
--    assert.equal(engine.busy, false);
--  } finally {
--    await engine.stop();
--  }
--});
-+test("engine: one prompt, one turn, text and usage come back", () => withEngine({}, async ({ engine }) => {
-+  const r = await engine.prompt("hello");
-+  assert.equal(r.text, "echo: hello");
-+  assert.deepEqual(r.usage, { input: 3, output: 2 });
-+  await idle(engine);
-+  assert.equal(engine.busy, false);
-+}));
- 
--test("engine: a prompt while streaming is held until pi settles, then sent as its own run, and answered in order", async () => {
--  const { engine, commands } = start(makeRoot());
--  const first = engine.prompt("slow 150");
--  await new Promise((r) => setTimeout(r, 20));
-+test("engine: a prompt while streaming is held until pi settles, then sent as its own run, and answered in order", () => withEngine({}, async ({ engine, commands }) => {
-+  const first = engine.prompt("slow 300");
-   assert.equal(engine.busy, true);
-   const second = engine.prompt("second");
-   assert.equal(engine.pendingCount, 2);
--  await new Promise((r) => setTimeout(r, 20));
--  assert.equal(commands().filter((c) => c.type === "prompt").length, 1, "the second prompt is not sent while pi is busy");
-+  const prompted = () => commands().filter((c) => c.type === "prompt");
-+  assert.ok(await until(() => prompted().length > 0), "the first prompt reached pi");
-+  assert.equal(prompted().length, 1, "the second prompt is not sent while pi is busy");
-+  // The fake refuses a prompt without streamingBehavior while it runs one, so
-+  // an answered second prompt also proves it was not sent early.
-   const [r1, r2] = await Promise.all([first, second]);
-   assert.equal(r1.text, "slow reply");
-   assert.equal(r2.text, "echo: second");
--  const prompts = commands().filter((c) => c.type === "prompt");
-+  const prompts = prompted();
-   assert.equal(prompts.length, 2);
-   // Never a pi follow-up: pi would fold it into the first run and close both
-   // answers with one agent_end (the live loss of 2026-09-17).
-@@ -117,11 +122,9 @@ test("engine: a prompt while streaming is held until pi settles, then sent as it
-   assert.equal(prompts[1].streamingBehavior, undefined);
-   await idle(engine);
-   assert.equal(engine.busy, false);
--  await engine.stop();
--});
-+}));
- 
--test("engine: a held prompt that times out before pi settles fails on its own and is never sent", async () => {
--  const { engine, commands } = start(makeRoot());
-+test("engine: a held prompt that times out before pi settles fails on its own and is never sent", () => withEngine({}, async ({ engine, commands }) => {
-   const first = engine.prompt("slow 200");
-   await new Promise((r) => setTimeout(r, 20));
-   await assert.rejects(engine.prompt("late one", { timeoutMs: 50 }), (e) => e.details.code === "timeout" && /waiting for the engine/.test(e.message));
-@@ -130,50 +133,85 @@ test("engine: a held prompt that times out before pi settles fails on its own an
-   await idle(engine);
-   assert.deepEqual(commands().filter((c) => c.type === "prompt").map((c) => c.message), ["slow 200"]);
-   assert.deepEqual(commands().filter((c) => c.type === "abort"), [], "a held turn is not aborted; pi never had it");
--  await engine.stop();
--});
-+}));
- 
--test("engine: timeout sends abort and fails only that turn; the process stays", async () => {
--  const { engine, commands, logs } = start(makeRoot());
-+test("engine: timeout sends abort and fails only that turn; the process stays", () => withEngine({}, async ({ engine, commands, logs }) => {
-   await assert.rejects(engine.prompt("slow 5000", { timeoutMs: 100 }), (err) => err.details.code === "timeout");
-   assert.ok(await until(() => commands().some((c) => c.type === "abort")), "abort reached pi");
-   assert.ok(logs.some((l) => /timed out/.test(l)));
-   const r = await engine.prompt("again");
-   assert.equal(r.text, "echo: again");
--  await engine.stop();
-+}));
-+
-+test("engine: tool events from a run that outlived its timeout never land in the next prompt's record", () => withEngine({}, async ({ engine }) => {
-+  await assert.rejects(engine.prompt("late 200", { timeoutMs: 40 }), (err) => err.details.code === "timeout");
-+  const r = await engine.prompt("after late");
-+  assert.equal(r.text, "echo: after late");
-+  assert.deepEqual(r.tools, [], "the dead run's read is not this prompt's evidence");
-+  assert.equal(r.turns, 1, "the dead run's turns are not counted here");
-+}));
-+
-+// The turn timer is fired by hand, before the engine has read any event from
-+// pi, so the timed-out run is still pi's and state.busy is still false when
-+// the next prompt arrives. Under load a real timer does the same.
-+const TURN_MS = 60000;
-+const manualTurnTimer = (fire) => ({
-+  setTimeoutImpl: (fn, ms) => (ms === TURN_MS ? fire.push(fn) : setTimeout(fn, ms)),
-+  clearTimeoutImpl: (id) => { if (typeof id !== "number") clearTimeout(id); },
- });
- 
--test("engine: tool events from a run that outlived its timeout never land in the next prompt's record", async () => {
--  const { engine } = start(makeRoot());
--  try {
--    await assert.rejects(engine.prompt("late 200", { timeoutMs: 40 }), (err) => err.details.code === "timeout");
--    const r = await engine.prompt("after late");
-+test("engine: a prompt after a turn that timed out before its agent_start waits for pi to settle instead of being refused", () => {
-+  const fire = [];
-+  return withEngine(manualTurnTimer(fire), async ({ engine, commands }) => {
-+    const late = engine.prompt("late 100", { timeoutMs: TURN_MS });
-+    fire.shift()();
-+    assert.equal(engine.busy, true, "pi is still running the prompt that timed out");
-+    const next = engine.prompt("after late", { timeoutMs: 5000 });
-+    assert.equal(engine.pendingCount, 1, "only the new prompt is live");
-+    await assert.rejects(late, (err) => err.details.code === "timeout");
-+    const r = await next;
-     assert.equal(r.text, "echo: after late");
-     assert.deepEqual(r.tools, [], "the dead run's read is not this prompt's evidence");
--    assert.equal(r.turns, 1, "the dead run's turns are not counted here");
--  } finally {
--    await engine.stop();
--  }
-+    assert.equal(r.turns, 1);
-+    assert.deepEqual(commands().map((c) => (c.type === "prompt" ? c.message : c.type)), ["late 100", "abort", "after late"]);
-+    await idle(engine);
-+    assert.equal(engine.busy, false);
-+  });
- });
- 
--test("engine: a malformed JSONL line fails the turn, not the process", async () => {
--  const { engine, logs } = start(makeRoot());
-+// "mute" is accepted and never run, so no agent_start, agent_end or settle
-+// ever comes for it. Unbounded, it would hold every later prompt.
-+test("engine: a timed-out turn pi never started holds the next prompt only for the abort grace, then leaves the queue", () => withEngine({ abortGraceMs: 150 }, async ({ engine, commands }) => {
-+  await assert.rejects(engine.prompt("mute", { timeoutMs: 50 }), (err) => err.details.code === "timeout");
-+  assert.equal(engine.busy, true, "pi might still be running it");
-+  const started = Date.now();
-+  const r = await engine.prompt("after mute", { timeoutMs: 5000 });
-+  assert.equal(r.text, "echo: after mute");
-+  assert.ok(Date.now() - started >= 100, "held for the grace, not sent at once");
-+  assert.deepEqual(commands().map((c) => (c.type === "prompt" ? c.message : c.type)), ["mute", "abort", "after mute"]);
-+  await idle(engine);
-+  assert.equal(engine.busy, false);
-+}));
-+
-+test("engine: a malformed JSONL line fails the turn, not the process", () => withEngine({}, async ({ engine, logs }) => {
-   await assert.rejects(engine.prompt("garbage"), (err) => err.details.code === "engine-protocol");
-   assert.ok(logs.some((l) => /malformed/.test(l)));
-   const r = await engine.prompt("still here");
-   assert.equal(r.text, "echo: still here");
--  await engine.stop();
--});
-+}));
- 
- test("engine: a turn that ends in error rejects with the error code; process exit fails pending turns", async () => {
--  const root = makeRoot();
-   let exited = null;
--  const { engine } = start(root, { onExit: (e) => (exited = e) });
--  await assert.rejects(engine.prompt("error"), (err) => err.details.code === "engine-error" && /fake provider error/.test(err.message));
--  const pending = engine.prompt("slow 5000");
--  await new Promise((r) => setTimeout(r, 20));
--  await engine.stop();
--  await assert.rejects(pending, (err) => err.details.code === "engine-down");
--  assert.ok(exited);
--  await assert.rejects(engine.prompt("x"), /not running/);
-+  const { engine } = start(makeRoot(), { onExit: (e) => (exited = e) });
-+  try {
-+    await assert.rejects(engine.prompt("error"), (err) => err.details.code === "engine-error" && /fake provider error/.test(err.message));
-+    const pending = engine.prompt("slow 5000");
-+    await new Promise((r) => setTimeout(r, 20));
-+    await engine.stop();
-+    await assert.rejects(pending, (err) => err.details.code === "engine-down");
-+    assert.ok(exited);
-+    await assert.rejects(engine.prompt("x"), /not running/);
-+  } finally {
-+    await engine.stop();
-+  }
- });
-diff --git a/packages/discord/tests/fake-pi.mjs b/packages/discord/tests/fake-pi.mjs
-index 94919306..ecc04e3b 100644
---- a/packages/discord/tests/fake-pi.mjs
-+++ b/packages/discord/tests/fake-pi.mjs
-@@ -8,6 +8,7 @@
- //                 then a second turn that answers "read <n> file(s)"
- //   "toolonly"    a run whose only turn calls a tool and never answers
- //   "retry"       an agent_end with willRetry, then the real answer
-+//   "mute"        accept the prompt and emit nothing, staying idle
- //   "late <ms>"   ignore abort; after <ms> emit a tool pair and a tool turn,
- //                 then answer "late reply", like a run that outlives its
- //                 client-side timeout
-@@ -30,6 +31,7 @@ function assistant(text, stopReason = "stop") {
- }
- 
- function run(text) {
-+  if (text === "mute") return;
-   busy = true;
-   out({ type: "agent_start" });
-   out({ type: "turn_start" });

agents/darkwing/work/discord-engine-busy/r2-manifest.sha256

@@ -1,3 +0,0 @@
-77077b7fbd5a933ffd352094eb073227c299ba47b7aea52d4e60fdc55cc7101e  packages/discord/src/engine-pi.mjs
-47a998179c6eb46827f43ab2c6b0f6b062ef94fb47da402cfb9af8c4f180f38e  packages/discord/tests/engine.test.mjs
-a8e54cc3f4b670eef2c06755b63e9e6bfeb44b1b1efde3aca9bfaa91583c3ef3  packages/discord/tests/fake-pi.mjs

agents/darkwing/work/discord-engine-busy/r2.patch

@@ -1,651 +0,0 @@
-diff --git a/packages/discord/src/engine-pi.mjs b/packages/discord/src/engine-pi.mjs
-index 5c8fd0a9..46ef1f88 100644
---- a/packages/discord/src/engine-pi.mjs
-+++ b/packages/discord/src/engine-pi.mjs
-@@ -16,9 +16,14 @@
- // from `tool_execution_start`/`tool_execution_end` into the result so the
- // turn record shows what was read. An `agent_end` with `willRetry` is not
- // the end of the run. A timeout sends `abort` and fails that turn; the
--// process stays. A malformed JSONL line from pi fails the current turn (its
--// outcome is now unknowable) and the process stays. Process exit fails
--// every pending turn and is reported through `onExit`.
-+// process stays. The failed turn holds later prompts back until its
-+// agent_end or a settle. If pi has not started it within ABORT_GRACE_MS, the
-+// engine stops pi instead of sending again: pi's events carry no prompt id,
-+// so a late run of the failed prompt would be taken for the next one's. A
-+// run pi did start holds later prompts until it ends, as any run does. A
-+// malformed JSONL line from pi fails the current turn (its outcome is now
-+// unknowable) and the process stays. Process exit fails every pending turn
-+// and is reported through `onExit`.
- //
- // Framing follows pi's RPC doc: split on "\n" only, strip a trailing "\r".
- // Node readline is not used because it also splits on U+2028/U+2029.
-@@ -64,31 +69,67 @@ export function assistantText(message) {
-     .trim();
- }
- 
-+// How long a turn that failed here (timeout, protocol error) may wait for
-+// pi's agent_start before the engine stops pi. Without a bound, a prompt pi
-+// accepted but never ran would hold every later prompt until restart.
-+export const ABORT_GRACE_MS = 30000;
-+
- export function createEngine({
-   command, args, cwd, env = {},
-   spawn = nodeSpawn,
-   setTimeoutImpl = globalThis.setTimeout, clearTimeoutImpl = globalThis.clearTimeout,
-   log = () => {},
-   onExit = () => {},
-+  abortGraceMs = ABORT_GRACE_MS,
- } = {}) {
-   if (typeof command !== "string" || command.length === 0) throw new DiscordError("engine: command required", 1);
-   if (!Array.isArray(args)) throw new DiscordError("engine: args required", 1);
- 
-   // pending: prompts sent to pi, oldest first. held: prompts waiting for pi
-   // to settle before they are sent, oldest first.
--  const state = { child: null, buffer: "", pending: [], held: [], responses: new Map(), nextId: 1, busy: false, exited: null };
-+  // wedged: set when the engine gave up on pi and is stopping it. Nothing
-+  // is sent to that child again.
-+  const state = { child: null, buffer: "", pending: [], held: [], responses: new Map(), nextId: 1, busy: false, exited: null, wedged: false };
- 
-   // A turn that fails on the client side (timeout, protocol error) stays in
-   // the pending queue, marked done, until pi's own turn_end for it arrives.
--  // Otherwise that turn_end would be attributed to the next prompt.
-+  // Otherwise that turn_end would be attributed to the next prompt. It holds
-+  // later prompts back; if pi has not started it within abortGraceMs, the
-+  // engine stops pi.
-   function failTurn(turn, code, message) {
-     if (turn.done) return;
-     turn.done = true;
-     if (turn.timer !== null) clearTimeoutImpl(turn.timer);
-     turn.timer = null;
-+    if (state.pending.includes(turn)) {
-+      turn.grace = setTimeoutImpl(() => {
-+        turn.grace = null;
-+        // A run pi started keeps its place until its agent_end or a settle.
-+        if (state.busy) return;
-+        // No agent_start yet. Pi may never run this prompt, or its events
-+        // may still be on the way; with no prompt id in them, nothing sent
-+        // now could be told apart from it. Stop pi: held prompts fail, and
-+        // the exit fails the rest and reaches onExit.
-+        log(`engine: no agent_start ${abortGraceMs} ms after a failed turn; stopping pi`);
-+        wedge();
-+      }, abortGraceMs);
-+    }
-     turn.reject(new DiscordError(message, 1, { code }));
-   }
- 
-+  function wedge() {
-+    if (state.wedged || state.exited !== null) return;
-+    state.wedged = true;
-+    for (const h of state.held.splice(0)) failTurn(h.turn, "engine-wedged", "engine stopped: pi did not start an aborted turn");
-+    stopChild();
-+  }
-+
-+  // Call when a turn leaves the pending queue.
-+  function release(turn) {
-+    if (turn.grace !== null) clearTimeoutImpl(turn.grace);
-+    turn.grace = null;
-+  }
-+
-   function settleTurn(turn, value) {
-     if (turn.done) return;
-     turn.done = true;
-@@ -99,7 +140,10 @@ export function createEngine({
- 
-   function failAll(code, message) {
-     const pending = state.pending.splice(0);
--    for (const t of pending) failTurn(t, code, message);
-+    for (const t of pending) {
-+      release(t);
-+      failTurn(t, code, message);
-+    }
-     for (const h of state.held.splice(0)) failTurn(h.turn, code, message);
-     for (const [, r] of state.responses) r.reject(new DiscordError(message, 1, { code }));
-     state.responses.clear();
-@@ -174,6 +218,7 @@ export function createEngine({
-       // Attribute the run to the head even if it failed client-side, so the
-       // next prompt's agent_end is not taken for this one.
-       const run = state.pending.shift();
-+      if (run) release(run);
-       if (!run || run.done) return;
-       const messages = Array.isArray(event.messages) ? event.messages.filter((m) => m && m.role === "assistant") : [];
-       const message = messages.length > 0 ? messages[messages.length - 1] : run.last;
-@@ -193,13 +238,14 @@ export function createEngine({
-       // this settle and still has no agent_end will never get one: fail it now
-       // instead of waiting for its timeout. Turns whose prompt response has
-       // not arrived yet belong to a later run and stay.
-+      const dropped = [];
-       const keep = [];
--      for (const t of state.pending) {
--        if (t.done) continue;
--        if (t.accepted) failTurn(t, "engine-settled-without-turn", "engine settled without answering this prompt");
--        else keep.push(t);
--      }
-+      for (const t of state.pending) (t.done || t.accepted ? dropped : keep).push(t);
-       state.pending = keep;
-+      for (const t of dropped) {
-+        release(t);
-+        failTurn(t, "engine-settled-without-turn", "engine settled without answering this prompt");
-+      }
-       sendHeld();
-     }
-   }
-@@ -214,21 +260,29 @@ export function createEngine({
-       // Never accepted: pi will not emit a turn_end for it, so remove it.
-       const i = state.pending.indexOf(turn);
-       if (i !== -1) state.pending.splice(i, 1);
-+      release(turn);
-       failTurn(turn, (err.details && err.details.code) || "engine-refused", err.message);
-       sendHeld();
-     });
-   }
- 
-+  // Pi is busy from our side while any sent prompt is still queued, even one
-+  // that already failed here: a turn that timed out before its agent_start
-+  // was read leaves state.busy false while pi runs it, and sending then would
-+  // be refused as streaming. It leaves the queue on its agent_end, on a
-+  // settle, on a refused send, or at process exit.
-+  const engineBusy = () => state.busy || state.pending.length > 0;
-+
-   // After a settle (or a refused send) the oldest held prompt goes out.
-   function sendHeld() {
--    if (state.exited !== null) return;
--    if (state.busy || state.pending.some((t) => !t.done)) return;
-+    if (state.exited !== null || state.wedged) return;
-+    if (engineBusy()) return;
-     const next = state.held.shift();
-     if (next) send(next.turn, next.command);
-   }
- 
-   function write(command) {
--    if (!state.child || state.exited !== null) throw new DiscordError("engine is not running", 1, { code: "engine-down" });
-+    if (!state.child || state.exited !== null || state.wedged) throw new DiscordError("engine is not running", 1, { code: "engine-down" });
-     state.child.stdin.write(JSON.stringify(command) + "\n");
-   }
- 
-@@ -245,6 +299,30 @@ export function createEngine({
-     });
-   }
- 
-+  function stopChild({ graceMs = 5000 } = {}) {
-+    const child = state.child;
-+    if (!child || state.exited !== null) return Promise.resolve(state.exited);
-+    return new Promise((resolve) => {
-+      const timer = setTimeoutImpl(() => {
-+        try {
-+          child.kill("SIGKILL");
-+        } catch {
-+          // already gone
-+        }
-+      }, graceMs);
-+      child.once("exit", () => {
-+        clearTimeoutImpl(timer);
-+        resolve(state.exited);
-+      });
-+      try {
-+        child.stdin.end();
-+        child.kill("SIGTERM");
-+      } catch {
-+        // already gone
-+      }
-+    });
-+  }
-+
-   return {
-     start() {
-       if (state.child) throw new DiscordError("engine already started", 1);
-@@ -281,7 +359,7 @@ export function createEngine({
-     // with DiscordError carrying details.code for the turn record.
-     prompt(text, { timeoutMs = 180000 } = {}) {
-       if (typeof text !== "string" || text.length === 0) throw new DiscordError("prompt text required", 1);
--      const turn = { resolve: null, reject: null, timer: null, done: false, accepted: false, tools: new Map(), turns: 0, last: null };
-+      const turn = { resolve: null, reject: null, timer: null, grace: null, done: false, accepted: false, tools: new Map(), turns: 0, last: null };
-       const done = new Promise((resolve, reject) => {
-         turn.resolve = resolve;
-         turn.reject = reject;
-@@ -306,44 +384,24 @@ export function createEngine({
-         }
-         failTurn(turn, "timeout", `turn timed out after ${timeoutMs} ms`);
-       }, timeoutMs);
--      if (state.exited !== null) {
-+      if (state.exited !== null || state.wedged) {
-         failTurn(turn, "engine-down", "engine is not running");
-         return done;
-       }
--      if (state.busy || state.pending.some((t) => !t.done) || state.held.length > 0) state.held.push({ turn, command });
-+      if (engineBusy() || state.held.length > 0) state.held.push({ turn, command });
-       else send(turn, command);
-       return done;
-     },
- 
-     get busy() {
--      return state.busy || state.pending.some((t) => !t.done) || state.held.length > 0;
-+      return engineBusy() || state.held.length > 0;
-     },
-     get pendingCount() {
-       return state.pending.filter((t) => !t.done).length + state.held.length;
-     },
- 
--    stop({ graceMs = 5000 } = {}) {
--      const child = state.child;
--      if (!child || state.exited !== null) return Promise.resolve(state.exited);
--      return new Promise((resolve) => {
--        const timer = setTimeoutImpl(() => {
--          try {
--            child.kill("SIGKILL");
--          } catch {
--            // already gone
--          }
--        }, graceMs);
--        child.once("exit", () => {
--          clearTimeoutImpl(timer);
--          resolve(state.exited);
--        });
--        try {
--          child.stdin.end();
--          child.kill("SIGTERM");
--        } catch {
--          // already gone
--        }
--      });
-+    stop(options) {
-+      return stopChild(options);
-     },
-   };
- }
-diff --git a/packages/discord/tests/engine.test.mjs b/packages/discord/tests/engine.test.mjs
-index 59674d1e..f3bd7525 100644
---- a/packages/discord/tests/engine.test.mjs
-+++ b/packages/discord/tests/engine.test.mjs
-@@ -4,6 +4,8 @@ import { readFileSync } from "node:fs";
- import { join } from "node:path";
- import { createEngine, buildPiArgs, PI_FIXED_ARGS, TOOLS_EXTENSION, READONLY_TOOLS_EXTENSION, assistantText } from "../src/engine-pi.mjs";
- import { existsSync } from "node:fs";
-+import { EventEmitter } from "node:events";
-+import { PassThrough } from "node:stream";
- import { makeRoot } from "./helpers.mjs";
- 
- const fakePi = join(import.meta.dirname, "fake-pi.mjs");
-@@ -28,7 +30,20 @@ function start(root, extra = {}) {
-     log: (m) => logs.push(m), ...extra,
-   });
-   engine.start();
--  return { engine, logs, commands: () => readFileSync(logPath, "utf8").trim().split("\n").filter(Boolean).map((l) => JSON.parse(l)) };
-+  // The fake creates its log on the first command; until then there are none.
-+  const commands = () => (existsSync(logPath) ? readFileSync(logPath, "utf8").trim().split("\n").filter(Boolean).map((l) => JSON.parse(l)) : []);
-+  return { engine, logs, commands };
-+}
-+
-+// Every test stops its engine in finally: a fake pi left running after a
-+// failed assertion keeps the test file from exiting.
-+async function withEngine(extra, body) {
-+  const started = start(makeRoot(), extra);
-+  try {
-+    await body(started);
-+  } finally {
-+    await started.engine.stop();
-+  }
- }
- 
- test("engine: buildPiArgs carries the fixed flags, engine settings, session dir and prompt file", () => {
-@@ -56,8 +71,7 @@ test("engine: with tools, buildPiArgs turns pi's own tools off, loads the extens
-   assert.equal(rw[rw.indexOf("--tools") + 1], "list_dir,read_file,search,write_file,edit_file", "a writable root adds exactly the two write tools");
- });
- 
--test("engine: a run with tool turns settles once, on the answer, with every tool call in the result", async () => {
--  const { engine } = start(makeRoot());
-+test("engine: a run with tool turns settles once, on the answer, with every tool call in the result", () => withEngine({}, async ({ engine }) => {
-   const r = await engine.prompt("tools 3");
-   assert.equal(r.text, "read 3 file(s)");
-   assert.equal(r.turns, 2);
-@@ -71,45 +85,38 @@ test("engine: a run with tool turns settles once, on the answer, with every tool
-   assert.equal(plain.turns, 1);
-   await idle(engine);
-   assert.equal(engine.busy, false);
--  await engine.stop();
--});
-+}));
- 
--test("engine: a run that ends on a tool-only turn fails the prompt as empty; a retried run settles on the real end", async () => {
--  const { engine } = start(makeRoot());
-+test("engine: a run that ends on a tool-only turn fails the prompt as empty; a retried run settles on the real end", () => withEngine({}, async ({ engine }) => {
-   const r = await engine.prompt("toolonly");
-   assert.equal(r.text, "", "no text: the connector turns this into engine-empty");
-   assert.equal(r.tools.length, 1);
-   const again = await engine.prompt("retry");
-   assert.equal(again.text, "after retry");
--  await engine.stop();
--});
-+}));
- 
--test("engine: one prompt, one turn, text and usage come back", async () => {
--  const { engine } = start(makeRoot());
--  try {
--    const r = await engine.prompt("hello");
--    assert.equal(r.text, "echo: hello");
--    assert.deepEqual(r.usage, { input: 3, output: 2 });
--    await idle(engine);
--    assert.equal(engine.busy, false);
--  } finally {
--    await engine.stop();
--  }
--});
-+test("engine: one prompt, one turn, text and usage come back", () => withEngine({}, async ({ engine }) => {
-+  const r = await engine.prompt("hello");
-+  assert.equal(r.text, "echo: hello");
-+  assert.deepEqual(r.usage, { input: 3, output: 2 });
-+  await idle(engine);
-+  assert.equal(engine.busy, false);
-+}));
- 
--test("engine: a prompt while streaming is held until pi settles, then sent as its own run, and answered in order", async () => {
--  const { engine, commands } = start(makeRoot());
--  const first = engine.prompt("slow 150");
--  await new Promise((r) => setTimeout(r, 20));
-+test("engine: a prompt while streaming is held until pi settles, then sent as its own run, and answered in order", () => withEngine({}, async ({ engine, commands }) => {
-+  const first = engine.prompt("slow 300");
-   assert.equal(engine.busy, true);
-   const second = engine.prompt("second");
-   assert.equal(engine.pendingCount, 2);
--  await new Promise((r) => setTimeout(r, 20));
--  assert.equal(commands().filter((c) => c.type === "prompt").length, 1, "the second prompt is not sent while pi is busy");
-+  const prompted = () => commands().filter((c) => c.type === "prompt");
-+  assert.ok(await until(() => prompted().length > 0), "the first prompt reached pi");
-+  assert.equal(prompted().length, 1, "the second prompt is not sent while pi is busy");
-+  // The fake refuses a prompt without streamingBehavior while it runs one, so
-+  // an answered second prompt also proves it was not sent early.
-   const [r1, r2] = await Promise.all([first, second]);
-   assert.equal(r1.text, "slow reply");
-   assert.equal(r2.text, "echo: second");
--  const prompts = commands().filter((c) => c.type === "prompt");
-+  const prompts = prompted();
-   assert.equal(prompts.length, 2);
-   // Never a pi follow-up: pi would fold it into the first run and close both
-   // answers with one agent_end (the live loss of 2026-09-17).
-@@ -117,11 +124,9 @@ test("engine: a prompt while streaming is held until pi settles, then sent as it
-   assert.equal(prompts[1].streamingBehavior, undefined);
-   await idle(engine);
-   assert.equal(engine.busy, false);
--  await engine.stop();
--});
-+}));
- 
--test("engine: a held prompt that times out before pi settles fails on its own and is never sent", async () => {
--  const { engine, commands } = start(makeRoot());
-+test("engine: a held prompt that times out before pi settles fails on its own and is never sent", () => withEngine({}, async ({ engine, commands }) => {
-   const first = engine.prompt("slow 200");
-   await new Promise((r) => setTimeout(r, 20));
-   await assert.rejects(engine.prompt("late one", { timeoutMs: 50 }), (e) => e.details.code === "timeout" && /waiting for the engine/.test(e.message));
-@@ -130,50 +135,177 @@ test("engine: a held prompt that times out before pi settles fails on its own an
-   await idle(engine);
-   assert.deepEqual(commands().filter((c) => c.type === "prompt").map((c) => c.message), ["slow 200"]);
-   assert.deepEqual(commands().filter((c) => c.type === "abort"), [], "a held turn is not aborted; pi never had it");
--  await engine.stop();
--});
-+}));
- 
--test("engine: timeout sends abort and fails only that turn; the process stays", async () => {
--  const { engine, commands, logs } = start(makeRoot());
-+test("engine: timeout sends abort and fails only that turn; the process stays", () => withEngine({}, async ({ engine, commands, logs }) => {
-   await assert.rejects(engine.prompt("slow 5000", { timeoutMs: 100 }), (err) => err.details.code === "timeout");
-   assert.ok(await until(() => commands().some((c) => c.type === "abort")), "abort reached pi");
-   assert.ok(logs.some((l) => /timed out/.test(l)));
-   const r = await engine.prompt("again");
-   assert.equal(r.text, "echo: again");
--  await engine.stop();
-+}));
-+
-+test("engine: tool events from a run that outlived its timeout never land in the next prompt's record", () => withEngine({}, async ({ engine }) => {
-+  await assert.rejects(engine.prompt("late 200", { timeoutMs: 40 }), (err) => err.details.code === "timeout");
-+  const r = await engine.prompt("after late");
-+  assert.equal(r.text, "echo: after late");
-+  assert.deepEqual(r.tools, [], "the dead run's read is not this prompt's evidence");
-+  assert.equal(r.turns, 1, "the dead run's turns are not counted here");
-+}));
-+
-+// The turn timer is fired by hand, before the engine has read any event from
-+// pi, so the timed-out run is still pi's and state.busy is still false when
-+// the next prompt arrives. Under load a real timer does the same.
-+const TURN_MS = 60000;
-+const manualTurnTimer = (fire) => ({
-+  setTimeoutImpl: (fn, ms) => (ms === TURN_MS ? fire.push(fn) : setTimeout(fn, ms)),
-+  clearTimeoutImpl: (id) => { if (typeof id !== "number") clearTimeout(id); },
- });
- 
--test("engine: tool events from a run that outlived its timeout never land in the next prompt's record", async () => {
--  const { engine } = start(makeRoot());
--  try {
--    await assert.rejects(engine.prompt("late 200", { timeoutMs: 40 }), (err) => err.details.code === "timeout");
--    const r = await engine.prompt("after late");
-+test("engine: a prompt after a turn that timed out before its agent_start waits for pi to settle instead of being refused", () => {
-+  const fire = [];
-+  return withEngine(manualTurnTimer(fire), async ({ engine, commands }) => {
-+    const late = engine.prompt("late 100", { timeoutMs: TURN_MS });
-+    fire.shift()();
-+    assert.equal(engine.busy, true, "pi is still running the prompt that timed out");
-+    const next = engine.prompt("after late", { timeoutMs: 5000 });
-+    assert.equal(engine.pendingCount, 1, "only the new prompt is live");
-+    await assert.rejects(late, (err) => err.details.code === "timeout");
-+    const r = await next;
-     assert.equal(r.text, "echo: after late");
-     assert.deepEqual(r.tools, [], "the dead run's read is not this prompt's evidence");
--    assert.equal(r.turns, 1, "the dead run's turns are not counted here");
--  } finally {
--    await engine.stop();
--  }
-+    assert.equal(r.turns, 1);
-+    assert.deepEqual(commands().map((c) => (c.type === "prompt" ? c.message : c.type)), ["late 100", "abort", "after late"]);
-+    await idle(engine);
-+    assert.equal(engine.busy, false);
-+  });
-+});
-+
-+// "mute" is accepted and never run, so no agent_start, agent_end or settle
-+// ever comes for it. Unbounded, it would hold every later prompt.
-+test("engine: when pi has not started a timed-out turn by the end of the abort grace, the engine stops pi and fails held prompts", async () => {
-+  let exited = null;
-+  await withEngine({ abortGraceMs: 150, onExit: (e) => (exited = e) }, async ({ engine, commands, logs }) => {
-+    await assert.rejects(engine.prompt("mute", { timeoutMs: 50 }), (err) => err.details.code === "timeout");
-+    assert.equal(engine.busy, true, "pi might still be running it");
-+    const started = Date.now();
-+    await assert.rejects(engine.prompt("after mute", { timeoutMs: 5000 }), (err) => err.details.code === "engine-wedged");
-+    assert.ok(Date.now() - started >= 100, "held for the grace, not failed at once");
-+    await assert.rejects(engine.prompt("later"), (err) => err.details.code === "engine-down");
-+    assert.ok(await until(() => exited !== null), "pi exits and onExit hears of it");
-+    assert.ok(logs.some((l) => /stopping pi/.test(l)));
-+    assert.deepEqual(commands().map((c) => (c.type === "prompt" ? c.message : c.type)), ["mute", "abort"]);
-+  });
-+});
-+
-+// Rocko's 6b R1 case: pi is stuck before agent_start, then runs the old
-+// prompt and only afterwards reads the next one. The events carry no prompt
-+// id, so a prompt sent after the grace would get the old run's answer.
-+test("engine: a timed-out turn pi starts only after the grace never answers a later prompt", async () => {
-+  let exited = null;
-+  await withEngine({ abortGraceMs: 150, onExit: (e) => (exited = e) }, async ({ engine, commands }) => {
-+    await assert.rejects(engine.prompt("stall 400", { timeoutMs: 50 }), (err) => err.details.code === "timeout");
-+    await assert.rejects(engine.prompt("after stall", { timeoutMs: 5000 }), (err) => err.details.code === "engine-wedged");
-+    assert.ok(await until(() => exited !== null), "pi exits and onExit hears of it");
-+    await new Promise((res) => setTimeout(res, 400));
-+    assert.ok(!commands().some((c) => c.message === "after stall"), "nothing was sent after the grace");
-+  });
- });
- 
--test("engine: a malformed JSONL line fails the turn, not the process", async () => {
--  const { engine, logs } = start(makeRoot());
-+// The same case in memory, after Rocko's reproducer: the old run's events
-+// arrive after the grace while pi is still exiting. They land on the failed
-+// turn, nothing more is written to pi, and only the exit ends the engine.
-+// Pi's response to the old prompt comes either before its timeout or only
-+// with the late events.
-+for (const lateResponse of [false, true]) test(`engine: late events of a run past its grace, before pi exits, answer nothing and nothing more is sent (${lateResponse ? "late" : "early"} prompt response)`, async () => {
-+  const timers = [];
-+  const written = [];
-+  const kills = [];
-+  const child = new EventEmitter();
-+  child.stdout = new PassThrough();
-+  child.stderr = new PassThrough();
-+  child.stdin = { write: (s) => { written.push(JSON.parse(s)); return true; }, end: () => {} };
-+  child.kill = (signal) => { kills.push(signal); return true; };
-+  let exited = null;
-+  const engine = createEngine({
-+    command: "memory-only", args: [], spawn: () => child, abortGraceMs: 150, onExit: (e) => (exited = e),
-+    setTimeoutImpl: (fn, ms) => { const t = { fn, ms, active: true }; timers.push(t); return t; },
-+    clearTimeoutImpl: (t) => { t.active = false; },
-+  });
-+  const emit = (x) => child.stdout.write(JSON.stringify(x) + "\n");
-+  const fire = (ms) => { const t = timers.find((x) => x.ms === ms && x.active); assert.ok(t, `timer ${ms}`); t.active = false; t.fn(); };
-+  const message = (text) => ({ role: "assistant", content: [{ type: "text", text }], stopReason: "stop" });
-+  const tick = () => new Promise((res) => setImmediate(res));
-+  // Checked after a tick instead of awaited, so a regression fails here
-+  // rather than hanging on a promise nothing will settle.
-+  const outcome = (p) => {
-+    const o = { state: "pending", code: null, text: null };
-+    p.then((v) => Object.assign(o, { state: "resolved", text: v.text }), (e) => Object.assign(o, { state: "rejected", code: e.details && e.details.code }));
-+    return o;
-+  };
-+  engine.start();
-+  const first = engine.prompt("old", { timeoutMs: 50 });
-+  const accept = () => emit({ type: "response", id: written[0].id, command: "prompt", success: true });
-+  if (!lateResponse) accept();
-+  await tick();
-+  fire(50);
-+  await assert.rejects(first, (err) => err.details.code === "timeout");
-+  const next = outcome(engine.prompt("new", { timeoutMs: 2000 }));
-+  fire(150);
-+  await tick();
-+  assert.deepEqual(next, { state: "rejected", code: "engine-wedged", text: null });
-+  assert.deepEqual(kills, ["SIGTERM"]);
-+  if (lateResponse) accept();
-+  emit({ type: "agent_start" });
-+  emit({ type: "tool_execution_start", toolCallId: "old-call", toolName: "read_file", args: { root: "docs", path: "old.md" } });
-+  emit({ type: "tool_execution_end", toolCallId: "old-call", toolName: "read_file", result: { details: { root: "docs", path: "old.md", ok: true } } });
-+  emit({ type: "turn_end", message: message("OLD RUN ANSWER") });
-+  emit({ type: "agent_end", messages: [message("OLD RUN ANSWER")] });
-+  emit({ type: "agent_settled" });
-+  await tick();
-+  const after = outcome(engine.prompt("after settle", { timeoutMs: 2000 }));
-+  await tick();
-+  assert.deepEqual(after, { state: "rejected", code: "engine-down", text: null });
-+  assert.deepEqual(next, { state: "rejected", code: "engine-wedged", text: null }, "the old answer did not reach the new prompt");
-+  assert.deepEqual(written.map((c) => (c.type === "prompt" ? c.message : c.type)), ["old", "abort"], "no prompt reached pi after the grace");
-+  assert.equal(exited, null);
-+  fire(5000);
-+  assert.deepEqual(kills, ["SIGTERM", "SIGKILL"]);
-+  child.emit("exit", null, "SIGKILL");
-+  assert.deepEqual(exited, { code: null, signal: "SIGKILL" });
-+});
-+
-+test("engine: a timed-out run pi did start outlives the grace; the next prompt goes out when it ends", async () => {
-+  let exited = null;
-+  await withEngine({ abortGraceMs: 150, onExit: (e) => (exited = e) }, async ({ engine, commands }) => {
-+    await assert.rejects(engine.prompt("late 400", { timeoutMs: 50 }), (err) => err.details.code === "timeout");
-+    const r = await engine.prompt("after late", { timeoutMs: 5000 });
-+    assert.equal(r.text, "echo: after late");
-+    assert.deepEqual(r.tools, []);
-+    assert.equal(exited, null, "pi was not stopped");
-+    assert.deepEqual(commands().map((c) => (c.type === "prompt" ? c.message : c.type)), ["late 400", "abort", "after late"]);
-+  });
-+});
-+
-+test("engine: a malformed JSONL line fails the turn, not the process", () => withEngine({}, async ({ engine, logs }) => {
-   await assert.rejects(engine.prompt("garbage"), (err) => err.details.code === "engine-protocol");
-   assert.ok(logs.some((l) => /malformed/.test(l)));
-   const r = await engine.prompt("still here");
-   assert.equal(r.text, "echo: still here");
--  await engine.stop();
--});
-+}));
- 
- test("engine: a turn that ends in error rejects with the error code; process exit fails pending turns", async () => {
--  const root = makeRoot();
-   let exited = null;
--  const { engine } = start(root, { onExit: (e) => (exited = e) });
--  await assert.rejects(engine.prompt("error"), (err) => err.details.code === "engine-error" && /fake provider error/.test(err.message));
--  const pending = engine.prompt("slow 5000");
--  await new Promise((r) => setTimeout(r, 20));
--  await engine.stop();
--  await assert.rejects(pending, (err) => err.details.code === "engine-down");
--  assert.ok(exited);
--  await assert.rejects(engine.prompt("x"), /not running/);
-+  const { engine } = start(makeRoot(), { onExit: (e) => (exited = e) });
-+  try {
-+    await assert.rejects(engine.prompt("error"), (err) => err.details.code === "engine-error" && /fake provider error/.test(err.message));
-+    const pending = engine.prompt("slow 5000");
-+    await new Promise((r) => setTimeout(r, 20));
-+    await engine.stop();
-+    await assert.rejects(pending, (err) => err.details.code === "engine-down");
-+    assert.ok(exited);
-+    await assert.rejects(engine.prompt("x"), /not running/);
-+  } finally {
-+    await engine.stop();
-+  }
- });
-diff --git a/packages/discord/tests/fake-pi.mjs b/packages/discord/tests/fake-pi.mjs
-index 94919306..bf94c013 100644
---- a/packages/discord/tests/fake-pi.mjs
-+++ b/packages/discord/tests/fake-pi.mjs
-@@ -8,9 +8,13 @@
- //                 then a second turn that answers "read <n> file(s)"
- //   "toolonly"    a run whose only turn calls a tool and never answers
- //   "retry"       an agent_end with willRetry, then the real answer
-+//   "mute"        accept the prompt and emit nothing, staying idle
- //   "late <ms>"   ignore abort; after <ms> emit a tool pair and a tool turn,
- //                 then answer "late reply", like a run that outlives its
- //                 client-side timeout
-+//   "stall <ms>"  accept the prompt, then read nothing for <ms> (pi stuck
-+//                 before agent_start); then run it, answering "echo:
-+//                 stalled", and only then read what came in meanwhile
- //   anything else answer "echo: <text>" immediately
- // A prompt received while busy without streamingBehavior is refused, as pi
- // does. A prompt with streamingBehavior followUp is folded into the running
-@@ -30,6 +34,7 @@ function assistant(text, stopReason = "stop") {
- }
- 
- function run(text) {
-+  if (text === "mute") return;
-   busy = true;
-   out({ type: "agent_start" });
-   out({ type: "turn_start" });
-@@ -109,11 +114,16 @@ function run(text) {
- let current = null;
- 
- let buffer = "";
-+let stalled = false;
- process.stdin.setEncoding("utf8");
- process.stdin.on("data", (chunk) => {
-   buffer += chunk;
-+  drain();
-+});
-+
-+function drain() {
-   let idx;
--  while ((idx = buffer.indexOf("\n")) !== -1) {
-+  while (!stalled && (idx = buffer.indexOf("\n")) !== -1) {
-     const line = buffer.slice(0, idx);
-     buffer = buffer.slice(idx + 1);
-     if (!line) continue;
-@@ -125,7 +135,15 @@ process.stdin.on("data", (chunk) => {
-         continue;
-       }
-       out({ id: cmd.id, type: "response", command: "prompt", success: true });
--      if (busy) queue.push(cmd.message);
-+      const sm = /^stall (\d+)$/.exec(cmd.message);
-+      if (sm) {
-+        stalled = true;
-+        setTimeout(() => {
-+          run("stalled");
-+          stalled = false;
-+          drain();
-+        }, Number(sm[1]));
-+      } else if (busy) queue.push(cmd.message);
-       else run(cmd.message);
-     } else if (cmd.type === "abort") {
-       out({ id: cmd.id, type: "response", command: "abort", success: true });
-@@ -139,5 +157,5 @@ process.stdin.on("data", (chunk) => {
-       out({ id: cmd.id, type: "response", command: cmd.type, success: true, data: {} });
-     }
-   }
--});
-+}
- process.stdin.on("end", () => process.exit(0));

agents/darkwing/work/internal-development-bootstrap/live-smoke.json

@@ -1,25 +0,0 @@
-{
-  "observedAt": "2026-09-13T19:56:18.961314+00:00",
-  "issue": 1510,
-  "ownerAuthorizedLiveSmoke": true,
-  "researcher": [
-    {
-      "session": ".pi/state/researcher/sessions/2026-09-13T19-52-28-745Z_01a09c54-0b48-7154-addd-8fdce875aa4a.jsonl",
-      "entryId": "64b15fd5",
-      "timestamp": "2026-09-13T19:52:43.079Z",
-      "response": "RESEARCHER_NATIVE_SMOKE_OK",
-      "entrySha256": "d9cbaa5ea640d2b858a86b2fa24d3240b351d9383ffaee9721dd86fcd080c329"
-    }
-  ],
-  "rocko": {
-    "newLaunch": "refused by existing native launch lock",
-    "existingPid": 3707667,
-    "cwd": "/mnt/storage/src/mosaic-stack",
-    "nativeContextVerified": true,
-    "sonnetFlagVerified": true,
-    "socket": "mosaic-fleet",
-    "newModelResponseTested": false
-  },
-  "existingProcessesRestarted": false,
-  "homeLaunchersModified": false
-}

agents/darkwing/work/internal-development-bootstrap/r1-review.json

@@ -1,19 +0,0 @@
-{
-  "issue": 1510,
-  "candidate": "/tmp/internal-team-r1-i9t21yzr",
-  "manifestSha256": "23a27014ce6f04ce8187d8495b2efe62b814c3c7d041b027f99b1ec4d490709d",
-  "files": [
-    "AGENTS.md",
-    "agents/README.md",
-    "agents/researcher/SOUL.md",
-    "agents/researcher/CONTEXT.md",
-    "agents/researcher/README.md",
-    "agents/researcher/launch.sh",
-    "agents/researcher/validate-sessions.mjs",
-    "scripts/test-darkwing-launch.mjs",
-    "docs/plans/2026-09-13_internal-development-bootstrap.md"
-  ],
-  "tests": "node --test scripts/test-darkwing-launch.mjs scripts/test-rocko-launch.mjs",
-  "passed": 6,
-  "state": "ready for independent review"
-}

agents/darkwing/work/ledger-t3-header/README.md

@@ -1,77 +0,0 @@
-# Ledger: T3 header counts as agent (#1506), R1 candidate
-
-Sage assigned this on 2026-09-26 after commit A (af4203ca). Filbert reviews.
-Not committed.
-
-## Defect
-
-`messageKind` in `packages/ledger/src/ledger.mjs` knew only the tmux preamble
-`[host:session -> host:session]`. A prompt that opens with the T3 header
-`[from: sage (1ef1e4f8-…) -> to: filbert (9cb9731e-…) class=actionable]`
-counted as human, so Table 2's Human column and the human-per-closed ratio
-rise once seats talk over T3. DEFERRED Open entry "Ledger counts T3 agent
-messages as human".
-
-## Change
-
-- `messageKind` also matches the T3 header on the first line. The sender is the
-  `from:` role. `control-board` is board, any other role is agent. Anything short
-  of the full header stays human. That includes the header on a later line, a
-  leading space, a missing thread id, `class=` with capitals, `]` followed by a
-  non-space, and `From:` capitalized. The tmux branch is unchanged.
-- Two tests: a Table 2 fixture with two headered prompts and one plain prompt
-  for seat `bob`, expecting agent 2 and human 1. Also a direct classification table.
-- README counting rule names both forms.
-
-## Evidence
-
-- `node --test --test-reporter=tap packages/ledger/tests/`: 22/22 on the
-  working tree.
-- The same test file against HEAD's `ledger.mjs` (full `git archive HEAD` tree):
-  20/22. The two failures are the two new tests, so they catch the defect.
-  An earlier archive of `packages/ledger` alone also failed three gitea-helper
-  tests. Those tests need `scripts/gitea-api.sh`, which the partial archive left out.
-- Suites on the working tree: config 24, task 90, foundation 43, conductor 17,
-  release 14, auth 15, discord 63. None of them runs the ledger tests.
-
-## Frozen files
-
-`r1-manifest.sha256` holds the three file hashes; `r1.patch` is `git diff
-packages/ledger` at freeze time.
-
-## Known limit, not fixed here
-
-The fix changes zero current counts. Table 2 reads only
-`.pi/state/<seat>/sessions/*.jsonl`, and no file there contains a T3 header
-for any seat. Filbert checked this in R1:
-- `grep -rF '[from: ' .pi/state/*/sessions/` finds nothing.
-- The candidate `messageKind` gives Dewey 52 agent, 7 board and 9 human, and
-  Sage 193 agent and 72 human. That covers every user message in Pi logs
-  modified since 2026-09-20. Every non-human first line is the tmux form.
-- Filbert's three T3 messages to Dewey on 2026-09-26 are not in
-  `.pi/state/dewey`.
-
-Dewey's and Sage's Pi logs are current, but they only carry tmux traffic. T3
-traffic goes to the harness transcripts: Claude under `~/.claude/projects`,
-Codex under `~/.codex/sessions`. The ledger reads neither, so a T3-routed
-prompt to any seat counts nowhere, as agent or as human. The Human column
-can't see T3 traffic at all. For Darkwing and Filbert, whose newest Pi logs
-end 2026-09-14, and for Rocko, who has no Pi sessions directory, zero means an
-empty source, not zero human prompts. The fix is correct for a source that
-carries T3 headers. Sage asked for a brief on a read-only T3 thread source;
-Gate F waits on it.
-
-Filbert also found two misclassifications in older logs. Neither is touched
-here:
-- `[rev-code-02 -> dragon-lin:sage class=actionable]` has no host on the
-  sender, so it counts as human.
-- One Dewey prompt opens with a quote character before the tmux preamble, so
-  it counts as human.
-
-## Review
-
-Filbert, R1, 2026-09-26: approved the three frozen files. He verified the
-manifest and patch, got 22/22 on the tree and 20/22 against HEAD's source, and
-matched the regex to `docs/guides/T3-AGENT-COMMS.md`. He accepts the body on
-the header's line, which the tmux branch also allows. He corrected the
-known-limit text above.