kushi-agents 5.7.3 → 5.7.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "kushi-agents",
-  "version": "5.7.3",
+  "version": "5.7.4",
   "description": "Install Kushi — multi-source project evidence agent with Comprehensive Structured Capture (CSC) into weekly-only files across Email, Teams, OneNote, Loop, SharePoint, Meetings, CRM, ADO. Meetings retain a sibling verbatim/ audit folder. WorkIQ-only for M365 sources (Graph / m365_* FORBIDDEN as fallbacks; user-paste is first-class). Host-agnostic.",
   "type": "module",
   "bin": {

package/plugin/learnings/cross-cutting.md

@@ -4,6 +4,35 @@ Newest on top. Format defined in [`README.md`](./README.md). Use this file when
 
 ---
 
+### 2026-05-29 — Silent runners get killed by host watchdogs (GitHub Copilot Chat)
+
+**Symptom**: User invoked `discover.mjs` from inside GitHub Copilot Chat in VS Code. The runner appeared to hang for 60+ seconds, then the host announced "discover did not return completion output (it stayed silent/hung)" and gave up. Direct PowerShell invocation outside the host showed the same long pause but eventually returned.
+
+**Root cause**: `discover.mjs` runs 7 sequential WorkIQ calls (email, teams, meetings, onenote, sharepoint, crm, ado). Each call is 5–60s. The runner emitted **zero output** during the loop and only printed the JSON envelope at the end, so total silent time was up to 7 × 60s = 7 minutes. GitHub Copilot Chat's process watchdog interpreted the silence as a hung process and SIGKILLed it well before the runner finished.
+
+**Fix shipped (v5.7.4, 2026-05-29)**:
+
+1. **`discover.mjs` streams per-source progress to stderr** before and after each WorkIQ call:
+   ```
+   [discover] workiq: <path>
+   [discover] sources: email, teams, meetings, ... (timeout 90000ms each)
+   [discover] → email ...
+   [discover]   email ok in 12340ms (3 row(s))
+   [discover] → teams ...
+   ```
+   stdout stays reserved for the final JSON envelope (so machine consumers are unaffected).
+2. **Default `--timeout-ms` bumped from 60s to 90s** to absorb WorkIQ cold-start auth latency on the first call.
+3. **New regression test**: spawns discover with a fake workiq shim and asserts `[discover]` lines hit stderr.
+
+**Lesson — applies to every long-running runner**: any runner that takes >5s without emitting output is a host-watchdog timebomb. Two-channel discipline:
+
+- **stdout** = final structured envelope only (one JSON line, machine-readable).
+- **stderr** = streaming human-readable progress (every step, every long-running call).
+
+Adopting this convention now for all `plugin/runners/*.mjs` that loop over remote calls. Future runners should `process.stderr.write('[<runner>] step description\n')` before/after each external operation.
+
+---
+
 ### 2026-05-29 — `<auto>` placeholder check was blocking fresh installs
 
 **Symptom**: After a clean `npx kushi-agents@latest --all-hosts` install, the very first command users tried (`& '.\.kushi\lib\Get-KushiConfig.ps1' -Name 'm365-auth'`) threw a placeholder error citing `<auto>`. But `<auto>` is the explicit "Kushi will resolve this at runtime from identity / OS / WorkIQ" sentinel, by design — not an unfilled placeholder.

package/plugin/runners/discover.mjs

@@ -24,7 +24,7 @@ import { ask as workiqAsk, resolveWorkiqBin } from './lib/workiq.mjs';
 const ALL_SOURCES = ['email', 'teams', 'meetings', 'onenote', 'sharepoint', 'crm', 'ado'];
 
 function parseArgs(argv) {
-  const args = { force: false, dryRun: false, timeoutMs: 60_000, sources: null };
+  const args = { force: false, dryRun: false, timeoutMs: 90_000, sources: null };
   for (let i = 0; i < argv.length; i++) {
     const a = argv[i];
     if (a === '--project') args.project = argv[++i];
@@ -32,7 +32,7 @@ function parseArgs(argv) {
     else if (a === '--source') (args.sources ??= []).push(argv[++i]);
     else if (a === '--force') args.force = true;
     else if (a === '--dry-run') args.dryRun = true;
-    else if (a === '--timeout-ms') args.timeoutMs = Number(argv[++i]) || 60_000;
+    else if (a === '--timeout-ms') args.timeoutMs = Number(argv[++i]) || 90_000;
     else if (a === '--help' || a === '-h') args.help = true;
   }
   return args;
@@ -201,17 +201,27 @@ async function main() {
   let boundsDirty = false;
   let integDirty = false;
 
+  // Stream per-source progress to stderr so users see the runner is alive.
+  // stdout is reserved for the final JSON envelope.
+  const log = (msg) => process.stderr.write(`[discover] ${msg}\n`);
+  log(`workiq: ${workiqBin}`);
+  log(`sources: ${sourcesToRun.join(', ')} (timeout ${args.timeoutMs}ms each)`);
+
   for (const source of sourcesToRun) {
     const prompt = buildPrompt(source, path.basename(root));
     let rows = [];
     let asked = true;
     let skipReason = null;
+    const t0 = Date.now();
+    log(`→ ${source} ...`);
     try {
       const { blocks } = await workiqAsk(prompt, { bin: workiqBin, timeoutMs: args.timeoutMs });
       rows = rowsFromBlocks(blocks, source);
+      log(`  ${source} ok in ${Date.now() - t0}ms (${rows.length} row(s))`);
     } catch (e) {
       asked = true;
       skipReason = e.code || 'workiq-error';
+      log(`  ${source} ${skipReason} after ${Date.now() - t0}ms`);
     }
     const { boundariesPatch, integrationsPatch, accepted } = applyRows(source, rows, bounds, integ);
     if (boundariesPatch) {

package/plugin/runners/test/unit/discover.test.mjs

@@ -128,3 +128,25 @@ test('discover: --dry-run does not write files', async () => {
   // skipped path doesn't even check dry_run, so just confirm files_written is absent or empty
   assert.ok(!r.json.files_written || r.json.files_written.length === 0);
 });
+
+test('discover: streams [discover] progress to stderr (host-watchdog regression v5.7.4)', async () => {
+  // Without per-source stderr progress, hosts like GitHub Copilot Chat kill
+  // the process when stdout stays silent during slow workiq calls. Fix: emit
+  // `[discover] ...` lines to stderr so the host sees activity.
+  const bs = runRunner(BOOTSTRAP, ['--project', 'progress-proj', '--alias', 'alice']);
+  assert.equal(bs.code, 0, bs.stderr);
+
+  // Create a fake workiq shim that returns an empty response immediately.
+  const fakeBin = path.join(tmpRoot, process.platform === 'win32' ? 'fake-workiq.cmd' : 'fake-workiq');
+  const fakeBody = process.platform === 'win32' ? '@echo off\r\necho.\r\n' : '#!/bin/sh\necho ""\n';
+  await fs.writeFile(fakeBin, fakeBody);
+  if (process.platform !== 'win32') await fs.chmod(fakeBin, 0o755);
+
+  const r = runRunner(RUNNER, ['--project', 'progress-proj', '--alias', 'alice', '--source', 'email'], {
+    KUSHI_WORKIQ_BIN: fakeBin,
+  });
+  assert.equal(r.code, 0, r.stderr);
+  assert.match(r.stderr, /\[discover\] workiq:/, `expected workiq header on stderr, got: ${r.stderr}`);
+  assert.match(r.stderr, /\[discover\] sources: email/, `expected sources header on stderr, got: ${r.stderr}`);
+  assert.match(r.stderr, /\[discover\] → email/, `expected per-source progress on stderr, got: ${r.stderr}`);
+});

package/src/doctor.test.mjs

@@ -20,7 +20,7 @@ if (!fs.existsSync(TESTTMP)) fs.mkdirSync(TESTTMP, { recursive: true });
 
 function runDoctor(extraArgs = []) {
   const args = ['-NoProfile', '-File', doctorPs1, '-Repo', repoRoot, ...extraArgs];
-  const r = spawnSync('pwsh', args, { encoding: 'utf-8', timeout: 180_000 });
+  const r = spawnSync('pwsh', args, { encoding: 'utf-8', timeout: 300_000 });
   return { stdout: r.stdout || '', stderr: r.stderr || '', status: r.status ?? 1 };
 }