fix(console): D1 review fixes for the detached dev-owner supervisor (#1873)

Build review P2: the supervisor ignores a lost reply pipe and keeps
supervising. Security P2: a stop signal during start is held until the
start settles, and the CLI's start timeout signals the supervisor's
process group. Security P1 (part): console start --dev-local-owner
refuses inside an agent run. Tests cover review, confirm and reverse on
the org autonomy route; TOOLS.md and the webui README are corrected.

docs/TOOLS.md

@@ -333,7 +333,7 @@ For local development on a named instance only, `mosaic bus start BUSINESS --dev
 
 `mosaic console start --dev-local-owner --dev-insecure-bind ADDR --dev-allow-host NAME[,NAME...]` opens that dev owner Console to other machines, for UI testing from another device on a trusted LAN. It has no sign-in: anyone who can reach ADDR:PORT acts as the owner. Start, and `mosaic console status` while that Console runs, print `WARNING: dev owner Console listens on ADDR:PORT with no sign-in; anyone who can reach it acts as the owner.` on stderr. ADDR is `0.0.0.0` or an IPv4 address, and a specific address must also be one of the names. NAME is a host name or IPv4 address without a port; give several as one comma-separated value. The Console answers only `NAME:<bound port>` and the loopback names, and takes a write only when Origin is `http://` plus that same Host. A browser on such an http origin sends no Fetch Metadata, so for an allowed NAME a write with that exact Origin and no `Sec-Fetch-Site` counts as same-origin; a `Sec-Fetch-Site` the browser did send is still checked. Both flags are refused without `--dev-local-owner`, and `--dev-insecure-bind` is refused without a `--dev-allow-host`. SSO, preview and tester Consoles stay on loopback. The Console state file records `bind` and `allowHosts`. `mosaic console status` prints the state as JSON (`bind` is `127.0.0.1` for a loopback Console) and never moves a stale state file.
 
-`mosaic console restart` and `mosaic org reload ORG` are the Console's Settings controls (F3) from a terminal (U1, #1873). Each sends one request to the running Console named in the instance's console state, as the page sends it (the Console's own Host and Origin, `Sec-Fetch-Site: same-origin`, a JSON `{}` body), so the Console applies the same owner, `--send` and origin checks and writes the same action trail, `<dataRoot>/console/actions.jsonl`. Both refuse inside an agent run (exit 3) before reading anything, and refuse with exit 3 when no Console runs. `console restart` works only for a Console started in the foreground by `mosaic console start`, or by `mosaic console start --dev-local-owner --detach`, which leaves a detached supervisor (`packages/cli/src/console-supervisor.mjs`, its own session) running the Console as the foreground start does; its receipt adds the supervisor's `supervisor` pid, and its later warnings go to `console/supervisor-*.log` under the data root. Any other Console answers `not-supervised`; it waits up to 30 seconds for the relaunched Console to answer and prints `{"outcome":"restarted","from":OLD,"boot":NEW}`. `org reload ORG` restarts a running bus or starts a stopped one and prints the Console's done receipt. A request that left but got no answer is reported as outcome unknown (exit 1) and is never retried; check `mosaic console status`, `mosaic bus status ORG` and the trail. `mosaic bus restart ORG` remains the trail-less host verb.
+`mosaic console restart` and `mosaic org reload ORG` are the Console's Settings controls (F3) from a terminal (U1, #1873). Each sends one request to the running Console named in the instance's console state, as the page sends it (the Console's own Host and Origin, `Sec-Fetch-Site: same-origin`, a JSON `{}` body), so the Console applies the same owner, `--send` and origin checks and writes the same action trail, `<dataRoot>/console/actions.jsonl`. Both refuse inside an agent run (exit 3) before reading anything, and refuse with exit 3 when no Console runs. `console restart` works only for a Console started in the foreground by `mosaic console start`, or by `mosaic console start --dev-local-owner --detach`, which leaves a detached supervisor (`packages/cli/src/console-supervisor.mjs`, its own session) running the Console as the foreground start does; its receipt adds the supervisor's `supervisor` pid, and its later warnings go to `console/supervisor-*.log` under the data root. `mosaic console stop` stops that Console, and its supervisor then exits. `mosaic console start --dev-local-owner` refuses inside an agent run (exit 3), with or without `--detach`; a start without `--dev-local-owner` is unchanged. Any other Console answers `not-supervised`. `console restart` waits up to 30 seconds for the relaunched Console to answer and prints `{"outcome":"restarted","from":OLD,"boot":NEW}`. `org reload ORG` restarts a running bus or starts a stopped one and prints the Console's done receipt. A request that left but got no answer is reported as outcome unknown (exit 1) and is never retried; check `mosaic console status`, `mosaic bus status ORG` and the trail. `mosaic bus restart ORG` remains the trail-less host verb.
 
 The instance processing switch is separate from per-business launches off/on.
 `mosaic halt` durably stops managed processing across every business in the

packages/cli/src/cli.mjs

@@ -595,6 +595,8 @@ async function consoleCommand(rest, io, deps) {
   // org-lan-bind: refused here, before any config is read, unless both flags come with
   // --dev-local-owner. ADDR is 0.0.0.0 or an IPv4 address; NAME[,NAME...] has no ports.
   const devOwner = parsed.flags.has("--dev-local-owner"), bind = parsed.values["--dev-insecure-bind"] ?? null;
+  // D1: an agent run never starts a dev-local-owner Console (org-console.mjs consoleStart checks again).
+  if (devOwner) refuseInsideAgent(io.env);
   const names = Object.hasOwn(parsed.values, "--dev-allow-host") ? [parsed.values["--dev-allow-host"]] : [];
   let lan;
   try { lan = devLanOptions({ devOwner, bind, names }); } catch (err) { throw usage(err.message); }

packages/cli/src/console-supervisor.mjs

@@ -10,7 +10,15 @@ const { configDir, dataRoot, instance, port, bind, allowHosts } = JSON.parse(pro
 const warnings = [];
 let started = false;
 const warn = s => { if (started) process.stderr.write(s); else warnings.push(s); };
+// The starting CLI can be gone before the reply (Ctrl-C, a wrapper's timeout). The Console is
+// started and its state names it, so a reply nobody reads is dropped and supervision goes on.
+process.stdout.on('error', () => {});
 const reply = value => new Promise(resolve => process.stdout.write(`${JSON.stringify(value)}\n`, () => resolve()));
+// A stop signal during the start (the CLI's start timeout signals this process group) is kept and
+// acted on once the start settles, so a failed start still releases the start lock.
+let stopAsked = false;
+const early = () => { stopAsked = true; };
+for (const signal of ['SIGINT', 'SIGTERM']) process.on(signal, early);
 let receipt;
 try {
   receipt = await consoleStart({ configDir, dataRoot, instance, env: process.env }, { devLocalOwner: true, port, bind, allowHosts, warn });
@@ -23,7 +31,8 @@ try {
 started = true;
 // A stop signal ends supervision cleanly, as in the foreground (cli.mjs consoleCommand).
 const stop = () => receipt.stop();
-for (const signal of ['SIGINT', 'SIGTERM']) process.on(signal, stop);
+for (const signal of ['SIGINT', 'SIGTERM']) { process.off(signal, early); process.on(signal, stop); }
+if (stopAsked) stop();
 await reply({ receipt, warnings });
 const code = await receipt.done;
 if (code !== 0) process.stderr.write(`mosaic: console stopped unexpectedly (${code})\n`);

packages/cli/src/org-console.mjs

@@ -10,6 +10,7 @@ import { listOrgs, resolveOrg } from '../../organization/src/org-registry.mjs';
 import { cmdlineOf, startTimeOf } from './host.mjs';
 import { CliError } from './errors.mjs';
 import { devLanOptions } from './dev-local-owner.mjs';
+import { refuseInsideAgent } from './transport.mjs';
 
 const WEBUI = fileURLToPath(new URL('../../webui/src/cli.mjs', import.meta.url));
 const forbidden = new Set([7332, 7340, 7341, 33223, 17450, 17443, 44307]);
@@ -182,7 +183,16 @@ async function detachedSupervisor(ctx, { port, bind, allowHosts, warn }) {
   ]);
   clearTimeout(timer);
   child.stdout.destroy();
-  if (line === undefined) { child.kill('SIGTERM'); await done; fail(unclear); }
+  // The supervisor leads its own process group, with the Console it may have started: a timed-out
+  // start signals that group, then kills it if it is still there after 10 seconds.
+  if (line === undefined) {
+    const group = signal => { try { process.kill(-child.pid, signal); } catch { /* already gone */ } };
+    group('SIGTERM');
+    let grace; const exited = await Promise.race([done.then(() => true), new Promise(r => { grace = setTimeout(() => r(false), 10000); })]);
+    clearTimeout(grace);
+    if (!exited) { group('SIGKILL'); await done; }
+    fail(unclear);
+  }
   let reply = null;
   try { reply = JSON.parse(line); } catch { /* no reply */ }
   for (const w of Array.isArray(reply?.warnings) ? reply.warnings : []) if (typeof w === 'string') warn(w);
@@ -206,6 +216,9 @@ export async function consoleStart(ctx, { devLocalOwner = false, detach = false,
   if (!(port === 'auto' || (typeof port === 'string' && /^\d+$/.test(port) && value >= 1 && value <= 65535 && !forbidden.has(value)))) fail('invalid or forbidden console port');
   let lan;
   try { lan = devLanOptions({ devOwner: devLocalOwner, bind, names: allowHosts }); } catch (e) { throw new CliError(e.message, 4); }
+  // D1 fix round (security review P1): a dev-local-owner Console acts as the owner with no sign-in,
+  // so an agent run never starts one, detached or not. Other starts are unchanged.
+  if (devLocalOwner) refuseInsideAgent(ctx.env ?? process.env);
   if (detach && devLocalOwner) return detachedSupervisor(ctx, { port, bind, allowHosts, warn });
   const release = takeLock(ctx);
   let run, args;

packages/cli/tests/org-console.test.mjs

@@ -1,6 +1,6 @@
 import { test } from 'node:test';
 import assert from 'node:assert/strict';
-import { spawnSync } from 'node:child_process';
+import { spawn, spawnSync } from 'node:child_process';
 import { mkdtempSync, mkdirSync, writeFileSync, readdirSync, readFileSync, rmSync, existsSync, statSync } from 'node:fs';
 import { get, request } from 'node:http';
 import { join } from 'node:path';
@@ -237,7 +237,8 @@ test('a detached dev-owner console runs under a detached supervisor, so the owne
   // The page's request, as `mosaic console restart` sends it: 202, then a new boot on the same address.
   const restarted = await consoleRestart(ctx);
   assert.equal(restarted.outcome, 'restarted'); assert.equal(restarted.from, before.boot); assert.notEqual(restarted.boot, before.boot);
-  const saved = JSON.parse(readFileSync(stateFile, 'utf8'));
+  // The relaunched Console answers before the supervisor writes its state (relaunch), so wait for it.
+  const saved = await relaunchedState(ctx, result.pid);
   assert.notEqual(saved.pid, result.pid);
   assert.equal(saved.url, result.url);
   assert.equal(Number(statOf(saved.pid)[1]), result.supervisor);
@@ -272,6 +273,88 @@ test('a supervisor stopped by a signal stops its detached dev-owner console', {
   assert.equal(consoleStatus(ctx).running, false);
 });
 
+// relaunch() writes console.json after the new Console answers; read it once it names a new pid.
+async function relaunchedState(ctx, oldPid) {
+  const file = join(ctx.dataRoot, 'console/console.json');
+  for (const deadline = Date.now() + 15000; ; await new Promise(r => setTimeout(r, 50))) {
+    const saved = JSON.parse(readFileSync(file, 'utf8'));
+    if (saved.pid !== oldPid || Date.now() > deadline) return saved;
+  }
+}
+
+// D1 fix round (build review P2): the starting CLI can die before it reads the supervisor's reply
+// (Ctrl-C, a wrapper's timeout). The reply is dropped, and the supervisor keeps supervising.
+test('a supervisor whose starting CLI is gone keeps supervising, so Restart Console still relaunches', { timeout: 60000 }, async t => {
+  const { ctx } = await instance(t);
+  const child = spawn(process.execPath, [SUPERVISOR, JSON.stringify({ configDir: ctx.configDir, dataRoot: ctx.dataRoot, instance: 's1test', port: 'auto', bind: null, allowHosts: [] })],
+    { env: ctx.env, detached: true, stdio: ['ignore', 'pipe', 'pipe'] });
+  t.after(() => { try { process.kill(child.pid, 'SIGTERM'); } catch {} });
+  const exit = new Promise(resolve => child.once('exit', code => resolve(code)));
+  let stderr = ''; child.stderr.setEncoding('utf8').on('data', c => { stderr += c; });
+  child.stdout.destroy();
+  let status = consoleStatus(ctx);
+  for (const deadline = Date.now() + 30000; !status.running; status = consoleStatus(ctx)) {
+    if (Date.now() > deadline) assert.fail(`the console did not start: ${stderr}`);
+    await new Promise(r => setTimeout(r, 100));
+  }
+  // The reply follows the state write; give it time to meet the closed pipe.
+  await new Promise(r => setTimeout(r, 1000));
+  assert.equal(child.exitCode, null, stderr);
+  assert.equal(Number(statOf(status.pid)[1]), child.pid);
+  const restarted = await consoleRestart(ctx);
+  assert.equal(restarted.outcome, 'restarted');
+  const after = await relaunchedState(ctx, status.pid);
+  assert.notEqual(after.pid, status.pid); assert.equal(after.url, status.url);
+  assert.equal(Number(statOf(after.pid)[1]), child.pid);
+  assert.deepEqual(await consoleStop(ctx), { stopped: true, pid: after.pid });
+  assert.equal(await exit, 0);
+  assert.doesNotMatch(stderr, /EPIPE|Unhandled/);
+});
+
+// D1 fix round (security review P2): a stop signal while the supervisor holds the start lock is kept
+// until the start settles; then the Console stops, the lock is released and the supervisor exits.
+test('a supervisor signalled during its start stops the console it started and releases the start lock', { timeout: 60000 }, async t => {
+  const { ctx } = await instance(t);
+  const child = spawn(process.execPath, [SUPERVISOR, JSON.stringify({ configDir: ctx.configDir, dataRoot: ctx.dataRoot, instance: 's1test', port: 'auto', bind: null, allowHosts: [] })],
+    { env: ctx.env, detached: true, stdio: ['ignore', 'pipe', 'pipe'] });
+  t.after(() => { try { process.kill(-child.pid, 'SIGKILL'); } catch {} });
+  const exit = new Promise(resolve => child.once('exit', (code, signal) => resolve({ code, signal })));
+  let stdout = '', stderr = '';
+  child.stdout.setEncoding('utf8').on('data', c => { stdout += c; });
+  child.stderr.setEncoding('utf8').on('data', c => { stderr += c; });
+  const lock = join(ctx.dataRoot, 'console/start.lock');
+  for (const deadline = Date.now() + 20000; !existsSync(lock); await new Promise(r => setTimeout(r, 2))) {
+    if (Date.now() > deadline) assert.fail(`the supervisor never took the start lock: ${stderr}`);
+  }
+  process.kill(child.pid, 'SIGTERM');
+  assert.deepEqual(await exit, { code: 0, signal: null }, stderr);
+  assert.equal(existsSync(lock), false);
+  // As after a foreground stop by signal, the state file stays and reads stale, not running.
+  assert.equal(consoleStatus(ctx).running, false);
+  // The started Console was this supervisor's child, and it is gone.
+  const reply = JSON.parse(stdout.trim().split('\n')[0]);
+  assert.equal(existsSync(`/proc/${reply.receipt.pid}`), false);
+});
+
+// D1 fix round (security review P1): an agent run never starts a dev-local-owner Console.
+test('an agent run cannot start a dev-local-owner console; a start without it is unchanged', { timeout: 60000 }, async t => {
+  const { ctx, mosaic } = await instance(t);
+  const agent = { ...ctx, env: { ...ctx.env, MOSAIC_RUN_ID: 'run-1' } };
+  for (const detach of [true, false]) {
+    await assert.rejects(consoleStart(agent, { devLocalOwner: true, detach, port: 'auto', warn: () => {} }), e => e.exitCode === 3 && /only from a human shell/.test(e.message));
+  }
+  const agentCli = args => spawnSync(join(REPO, 'scripts/mosaic'), args, { env: { ...ctx.env, MOSAIC_AGENT_RUN: '1' }, encoding: 'utf8', timeout: 30000 });
+  const refused = agentCli(['console', 'start', '--dev-local-owner', '--detach', '--port', 'auto']);
+  assert.equal(refused.status, 3); assert.equal(refused.stdout, ''); assert.match(refused.stderr, /only from a human shell/);
+  const consoleDir = join(ctx.dataRoot, 'console');
+  assert.deepEqual(existsSync(consoleDir) ? readdirSync(consoleDir) : [], []);
+  assert.deepEqual(consoleStatus(ctx), { running: false });
+  const plain = agentCli(['console', 'start', '--detach', '--port', 'auto']);
+  assert.equal(plain.status, 0, plain.stderr);
+  assert.deepEqual(Object.keys(JSON.parse(plain.stdout)), ['url', 'pid']);
+  assert.equal((await consoleStop(ctx)).stopped, true);
+});
+
 // Without --dev-local-owner a detached start is what it was before D1: no supervisor, restart refused.
 test('a detached console without the dev owner has no supervisor, so it refuses a restart and stays up', { timeout: 60000 }, async t => {
   const { ctx, mosaic } = await instance(t);

packages/webui/README.md

@@ -22,7 +22,7 @@ baseline CLI.
 | Start, stop or restart a bus | `mosaic bus start <id>` / `mosaic bus stop <id>` / `mosaic bus restart <id>` | S6 integration hook |
 | Restart (or first start) an organization's bus | `mosaic bus restart <id>`, or `mosaic bus start <id>` for a bus never started | Enabled (F3): the owner's **Restart bus** control, offered where a page says the bus is down or a change needs a restart; trailed |
 | Restart the Console | Stop and run `mosaic console start` again | Enabled (F3) when `mosaic console start` runs it in the foreground, or with `--dev-local-owner --detach` under a detached supervisor (D1): Settings > Console; the owner only; trailed. Other Consoles answer `not-supervised` |
-| Save, review, confirm or reverse an organization's autonomy | None; console verbs `autonomy.set.console`, `autonomy.review.console`, `decision.confirm.console`, `decision.reverse.console` | Enabled (D1) on a `--dev-local-owner` Console: the owner only, on the selected organization's bus, behind a fence that passes those four verbs and nothing else |
+| Save, review, confirm or reverse an organization's autonomy | Save only: edit the autonomy block and run `mosaic org update ORG --from FILE --if-revision N`; no autonomy verb yet (readiness P1-3) | Enabled (D1) on a `--dev-local-owner` Console: the owner only, on the selected organization's bus, behind a fence that passes only the four console verbs `autonomy.set.console`, `autonomy.review.console`, `decision.confirm.console` and `decision.reverse.console` |
 | Create chat project linked to a configured repository | `mosaic project create <name> --repo <project> --org <id>` | Enabled |
 | Create a thread | `mosaic thread create <project> <name> --org <id>` | Planned enabled |
 | Open a thread | `mosaic thread open <thread> --org <id>` | Planned enabled |

packages/webui/tests/org-autonomy-write.test.mjs

@@ -42,6 +42,7 @@ export const consoleCall = o => async (verb, args, { subject }) => {
   record('console', o, verb, args, { subject });
   if (verb === 'autonomy.get.console') return { revision: 3, autonomy: { default: 'in' } };
   if (verb === 'autonomy.set.console') return { revision: 4, autonomy: args.autonomy, receipt: { action: 'autonomy.set', revision: 4 } };
+  if (['autonomy.review.console', 'decision.confirm.console', 'decision.reverse.console'].includes(verb)) return { answered: verb };
   throw new BusReadError('unavailable', 'fixture: ' + verb);
 };
 `);
@@ -109,6 +110,13 @@ test('the dev-local owner reads an organization autonomy page as writable and a
     { business: 'alpha', socket, verb: 'autonomy.get.console', args: {}, subject },
     { business: 'alpha', socket, verb: 'autonomy.set.console', args: AUTONOMY, subject },
   ]);
+  // Review, confirm and reverse go to the same organization bus as the owner (D1 fix round).
+  for (const [step, body, verb] of [['review', { decisionClass: 'merge' }, 'autonomy.review.console'], ['confirm', { id: 'd-1', confirm: true, reason: 'checked' }, 'decision.confirm.console'], ['reverse', { id: 'd-1' }, 'decision.reverse.console']]) {
+    const r = await call(port, `/api/orgs/alpha/autonomy/${step}`, { method: 'POST', body: JSON.stringify(body) });
+    assert.equal(r.status, 200, `${step}: ${JSON.stringify(r.body)}`);
+    assert.deepEqual(r.body, { answered: verb, writable: true });
+    assert.deepEqual(calls().filter(c => c.kind === 'console').at(-1), { kind: 'console', business: 'alpha', socket, verb, args: body, subject });
+  }
   // The write keeps the route's own checks: another site saves nothing.
   const before = calls().length;
   for (const headers of [{ origin: 'http://evil.example' }, { 'sec-fetch-site': 'cross-site' }]) {