forge-workflow 0.1.0-beta.4 → 0.1.0-beta.5

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (119) hide show
  1. package/AGENTS.md +14 -7
  2. package/CHANGELOG.md +43 -1
  3. package/README.md +6 -2
  4. package/bin/forge-cmd.js +20 -0
  5. package/bin/forge.js +16 -374
  6. package/docs/INDEX.md +1 -1
  7. package/docs/guides/BEADS_GITHUB_SYNC.md +2 -31
  8. package/docs/guides/MIGRATION.md +4 -4
  9. package/docs/guides/SETUP.md +16 -16
  10. package/docs/reference/COMMANDS.md +8 -5
  11. package/docs/reference/INSIGHTS_RECAP.md +9 -20
  12. package/docs/reference/RELEASE.md +5 -3
  13. package/docs/reference/TOOLCHAIN.md +8 -0
  14. package/docs/reference/protected-state-surfaces.md +4 -4
  15. package/docs/reference/shepherd.md +54 -25
  16. package/lefthook.yml +12 -0
  17. package/lib/activation/ensure-forge-home.js +33 -15
  18. package/lib/adapters/pr-state-adapter.js +344 -142
  19. package/lib/audit-evidence.js +71 -110
  20. package/lib/capped-jsonl-log.js +236 -0
  21. package/lib/commands/_registry.js +2 -2
  22. package/lib/commands/clean.js +196 -32
  23. package/lib/commands/dev.js +4 -33
  24. package/lib/commands/hooks.js +223 -25
  25. package/lib/commands/insights.js +8 -3
  26. package/lib/commands/merge.js +600 -40
  27. package/lib/commands/pr.js +1 -1
  28. package/lib/commands/preflight.js +11 -2
  29. package/lib/commands/prime.js +21 -8
  30. package/lib/commands/push.js +41 -51
  31. package/lib/commands/recall.js +60 -16
  32. package/lib/commands/recap.js +6 -1
  33. package/lib/commands/release.js +17 -2
  34. package/lib/commands/setup.js +191 -94
  35. package/lib/commands/shepherd.js +13 -1
  36. package/lib/commands/ship.js +22 -23
  37. package/lib/commands/skill.js +119 -11
  38. package/lib/commands/status.js +17 -1
  39. package/lib/commands/test.js +24 -34
  40. package/lib/commands/worktree.js +220 -42
  41. package/lib/core/runtime-graph.js +1 -1
  42. package/lib/doc-assertions.js +297 -0
  43. package/lib/existing-tdd-gate.js +253 -0
  44. package/lib/forge-context.js +1 -4
  45. package/lib/forge-issues.js +56 -32
  46. package/lib/git-defaults.js +56 -0
  47. package/lib/harness-capability-matrix.js +3 -3
  48. package/lib/hook-renderer.js +93 -4
  49. package/lib/insights.js +96 -80
  50. package/lib/kernel/backing-issue.js +14 -2
  51. package/lib/kernel/broker.js +16 -0
  52. package/lib/kernel/cli-broker-factory.js +12 -1
  53. package/lib/kernel/close-on-merge.js +154 -0
  54. package/lib/kernel/fs-class.js +42 -25
  55. package/lib/kernel/sqlite-driver.js +153 -29
  56. package/lib/lefthook-wiring.js +21 -1
  57. package/lib/memory/router.js +16 -1
  58. package/lib/memory-digest.js +47 -15
  59. package/lib/memory-recall-events.js +145 -0
  60. package/lib/memory-recall.js +71 -10
  61. package/lib/merge-rules.js +8 -4
  62. package/lib/npm-publish-workflow.js +272 -0
  63. package/lib/orientation.js +68 -43
  64. package/lib/plugin-catalog.js +14 -4
  65. package/lib/pr-bundle.js +5 -6
  66. package/lib/pr-monitor/journal.js +18 -2
  67. package/lib/pr-monitor/reconcile-executor.js +224 -41
  68. package/lib/pr-monitor/render-summary.js +196 -0
  69. package/lib/pr-monitor/shepherd-lease.js +10 -1
  70. package/lib/pr-monitor/watch-lifecycle.js +13 -1
  71. package/lib/pr-pull.js +33 -14
  72. package/lib/pr-shepherd.js +34 -8
  73. package/lib/preflight/gates.js +65 -18
  74. package/lib/preflight/runner.js +5 -0
  75. package/lib/project-memory.js +33 -1
  76. package/lib/protected-state-authority.js +305 -0
  77. package/lib/protected-state-surfaces.js +64 -44
  78. package/lib/release-readiness.js +51 -4
  79. package/lib/shell-utils.js +1 -1
  80. package/lib/skills-sync.js +6 -3
  81. package/lib/smart-merge.js +28 -4
  82. package/lib/symlink-utils.js +74 -26
  83. package/lib/upgrade-safety.js +39 -0
  84. package/lib/using-forge.js +19 -6
  85. package/package.json +6 -7
  86. package/scripts/doc-asserting-tests.js +158 -0
  87. package/scripts/lib/behavioral-eval-runner.js +310 -0
  88. package/scripts/lib/behavioral-eval-runtime.js +456 -0
  89. package/scripts/lib/eval-evidence.js +328 -0
  90. package/scripts/lib/eval-runner.js +81 -41
  91. package/scripts/lib/immutable-eval-corpus.js +309 -0
  92. package/scripts/lib/promotion-evidence-loader.js +94 -0
  93. package/scripts/lib/promotion-scorecard.js +314 -0
  94. package/scripts/npm-release-receipt.js +134 -0
  95. package/scripts/process-tree.js +761 -0
  96. package/scripts/protected-state-check.js +47 -22
  97. package/scripts/run-command-eval.js +29 -1
  98. package/scripts/sync-d20-audit.js +172 -0
  99. package/scripts/test-full-suite.js +249 -37
  100. package/scripts/test.js +176 -43
  101. package/skills/review/SKILL.md +4 -11
  102. package/skills/review/evals/scorecard.json +3 -3
  103. package/skills/rollback/SKILL.md +4 -11
  104. package/skills/rollback/evals/scorecard.json +3 -3
  105. package/skills/shepherd/SKILL.md +20 -14
  106. package/skills/shepherd/evals/scorecard.json +2 -2
  107. package/skills/ship/SKILL.md +4 -12
  108. package/skills/ship/evals/scorecard.json +3 -3
  109. package/skills/worktree/SKILL.md +6 -1
  110. package/skills/worktree/evals/scorecard.json +2 -2
  111. package/lib/beads-setup.js +0 -538
  112. package/lib/beads-sync-scaffold.js +0 -189
  113. package/lib/pat-setup.js +0 -207
  114. package/lib/pr-monitor/render-sticky.js +0 -206
  115. package/lib/pr-monitor/upsert-sticky.js +0 -169
  116. package/scripts/beads-context.sh +0 -577
  117. package/scripts/beads-migrate-to-dolt.sh +0 -7
  118. package/scripts/beads-upgrade-smoke.sh +0 -284
  119. package/scripts/lib/beads-migrate-to-dolt.mjs +0 -503
package/scripts/test.js CHANGED
@@ -4,15 +4,27 @@
4
4
  *
5
5
  * It runs only the tests affected by known changes when possible and falls back
6
6
  * to the full suite for package-level or unknown-file changes.
7
+ *
8
+ * Quick lane: `forge push --quick` sets FORGE_PUSH_LANE=quick on the `git push`
9
+ * it spawns, declaring the lint-only review-cycle lane. The pre-push entry point
10
+ * honors that declaration and skips the test run with a loud notice, so --quick
11
+ * is quick end to end instead of only in the step push.js controls. This is not
12
+ * a hook bypass: branch protection and lint still run, local validation
13
+ * (`--validate`) ignores the variable entirely, and CI runs the full matrix.
7
14
  */
8
15
 
9
- const { execFileSync: defaultExecFileSync, spawnSync: defaultSpawnSync } = require('node:child_process');
16
+ const { EventEmitter } = require('node:events');
17
+ const {
18
+ execFileSync: defaultExecFileSync,
19
+ spawn: defaultSpawn,
20
+ } = require('node:child_process');
10
21
  const fs = require('node:fs');
11
22
 
12
23
  const {
13
24
  getAffectedTestFiles,
14
25
  getChangedFiles,
15
26
  } = require('../lib/commands/test');
27
+ const { createProcessTree, signalExitCode } = require('./process-tree');
16
28
 
17
29
  const PACKAGE_LEVEL_PATHS = new Set([
18
30
  'package.json',
@@ -64,6 +76,21 @@ const DEFAULT_FULL_SUITE_TIMEOUT_MS = 10 * 60 * 1000;
64
76
  // Conventional shell exit code for a command terminated by a timeout.
65
77
  const TIMEOUT_EXIT_CODE = 124;
66
78
 
79
+ // Lane declaration set by `forge push --quick` on the spawned `git push`.
80
+ // Only this exact value opts into the lint-only lane.
81
+ const QUICK_LANE_ENV_VAR = 'FORGE_PUSH_LANE';
82
+ const QUICK_LANE_VALUE = 'quick';
83
+
84
+ /**
85
+ * Reports whether the environment declares the lint-only quick push lane.
86
+ *
87
+ * @param {NodeJS.ProcessEnv} [env=process.env] Environment to inspect.
88
+ * @returns {boolean} True when the quick lane is explicitly declared.
89
+ */
90
+ function isQuickPushLane(env = process.env) {
91
+ return env[QUICK_LANE_ENV_VAR] === QUICK_LANE_VALUE;
92
+ }
93
+
67
94
  /**
68
95
  * Reads and validates the FORGE_TEST_TIMEOUT_MS override, if any.
69
96
  *
@@ -134,9 +161,11 @@ function detectPackageManager() {
134
161
  function stripGitHookEnv(sourceEnv = process.env) {
135
162
  const env = { ...sourceEnv };
136
163
  for (const key of Object.keys(env)) {
137
- if (key === 'GIT_DIR' || key === 'GIT_WORK_TREE' || key === 'GIT_INDEX_FILE'
138
- || key === 'GIT_OBJECT_DIRECTORY' || key === 'GIT_ALTERNATE_OBJECT_DIRECTORIES'
139
- || key === 'GIT_QUARANTINE_PATH') {
164
+ const normalizedKey = key.toUpperCase();
165
+ if (normalizedKey === 'GIT_DIR' || normalizedKey === 'GIT_WORK_TREE'
166
+ || normalizedKey === 'GIT_INDEX_FILE' || normalizedKey === 'GIT_OBJECT_DIRECTORY'
167
+ || normalizedKey === 'GIT_ALTERNATE_OBJECT_DIRECTORIES'
168
+ || normalizedKey === 'GIT_QUARANTINE_PATH') {
140
169
  delete env[key];
141
170
  }
142
171
  }
@@ -317,31 +346,78 @@ function classifyPushTests(projectRoot, execFileSync = defaultExecFileSync) {
317
346
  *
318
347
  * @param {string} command Executable name.
319
348
  * @param {string[]} args Command arguments.
320
- * @param {import('node:child_process').SpawnSyncOptions} [options={}] Spawn options.
321
- * @param {typeof defaultSpawnSync} [spawnSync=defaultSpawnSync] Process runner.
322
- * @returns {number} Process exit status, or 1 when no status is reported.
349
+ * @param {import('node:child_process').SpawnOptions} [options={}] Spawn options.
350
+ * @param {typeof defaultSpawn} [spawn=defaultSpawn] Process runner.
351
+ * @returns {Promise<number>} Process exit status, or 1 when no status is reported.
323
352
  */
324
- function runCommand(command, args, options = {}, spawnSync = defaultSpawnSync) {
325
- const result = spawnSync(command, args, {
326
- stdio: 'inherit',
327
- shell: isWindows,
328
- ...options,
329
- });
330
-
331
- if (result.error) {
332
- if (result.error.code === 'ETIMEDOUT') {
333
- console.error('');
334
- console.error('Test lane exceeded its wall-clock ceiling and was terminated.');
335
- console.error('A single test likely hung (e.g. a spawned git/bash call that never returns).');
336
- console.error('Adjust the ceiling with FORGE_TEST_TIMEOUT_MS. Failing the run instead of');
337
- console.error('blocking the push. See issue 8aef79e8.');
338
- console.error('');
339
- return TIMEOUT_EXIT_CODE;
353
+ function runCommand(command, args, options = {}, spawn = defaultSpawn) {
354
+ const { processTree, timeout, killSignal: _killSignal, ...spawnOptions } = options;
355
+ return new Promise((resolve, reject) => {
356
+ let child;
357
+ let settled = false;
358
+ let timer = null;
359
+ const reservation = processTree?.reserveChild?.({ kind: 'test-lane', label: command }) || null;
360
+ let registered = false;
361
+ const release = () => {
362
+ if (registered) {
363
+ processTree?.unregisterChild?.(reservation);
364
+ registered = false;
365
+ }
366
+ };
367
+ const finish = (status) => {
368
+ if (settled) return;
369
+ settled = true;
370
+ release();
371
+ resolve(status ?? 1);
372
+ };
373
+ const fail = (error) => {
374
+ if (settled) return;
375
+ if (timer) clearTimeout(timer);
376
+ if (error?.code === 'ETIMEDOUT') {
377
+ processTree?.cleanup?.('SIGKILL');
378
+ console.error('');
379
+ console.error('Test lane exceeded its wall-clock ceiling and was terminated.');
380
+ console.error('A single test likely hung (e.g. a spawned git/bash call that never returns).');
381
+ console.error('Adjust the ceiling with FORGE_TEST_TIMEOUT_MS. Failing the run instead of');
382
+ console.error('blocking the push. See issue 8aef79e8.');
383
+ console.error('');
384
+ finish(TIMEOUT_EXIT_CODE);
385
+ return;
386
+ }
387
+ settled = true;
388
+ release();
389
+ reject(error);
390
+ };
391
+ try {
392
+ child = spawn(command, args, { stdio: 'inherit', shell: isWindows, ...spawnOptions });
393
+ } catch (error) {
394
+ fail(error);
395
+ return;
340
396
  }
341
- throw result.error;
342
- }
343
-
344
- return result.status ?? 1;
397
+ child.once?.('error', fail);
398
+ child.once?.('close', (status, signal) => {
399
+ if (timer) clearTimeout(timer);
400
+ finish(status ?? signalExitCode(signal));
401
+ });
402
+ if (reservation && typeof processTree?.registerChild === 'function') {
403
+ registered = true;
404
+ if (!processTree.registerChild(reservation, child)) {
405
+ processTree.abortChild?.(reservation, child);
406
+ release();
407
+ fail(new Error('test lane process could not be registered'));
408
+ return;
409
+ }
410
+ }
411
+ timer = timeout > 0 ? setTimeout(() => {
412
+ processTree?.cleanup?.('SIGKILL');
413
+ try { child.kill?.('SIGKILL'); } catch { /* best effort */ }
414
+ finish(TIMEOUT_EXIT_CODE);
415
+ }, timeout) : null;
416
+ if (!child.once) {
417
+ if (timer) clearTimeout(timer);
418
+ finish(child.status ?? 1);
419
+ }
420
+ });
345
421
  }
346
422
 
347
423
  /**
@@ -351,51 +427,92 @@ function runCommand(command, args, options = {}, spawnSync = defaultSpawnSync) {
351
427
  * @param {Object} [deps={}] Runtime dependencies for tests.
352
428
  * @returns {number} Exit status for the executed plan.
353
429
  */
354
- function runTestExecutionPlan(plan, deps = {}) {
355
- const spawnSync = deps.spawnSync || defaultSpawnSync;
430
+ async function runTestExecutionPlan(plan, deps = {}) {
431
+ const usesSpawnSyncAdapter = !deps.spawn && typeof deps.spawnSync === 'function';
432
+ const spawn = deps.spawn || (usesSpawnSyncAdapter ? (...args) => {
433
+ const child = new EventEmitter();
434
+ let result;
435
+ try {
436
+ result = deps.spawnSync(...args);
437
+ } catch (error) {
438
+ process.nextTick(() => child.emit('error', error));
439
+ return child;
440
+ }
441
+ child.pid = result?.pid || process.pid;
442
+ process.nextTick(() => {
443
+ if (result?.error) child.emit('error', result.error);
444
+ else child.emit('close', result?.status ?? 1, result?.signal);
445
+ });
446
+ return child;
447
+ } : defaultSpawn);
356
448
  const pkgManager = deps.pkgManager || detectPackageManager();
357
449
  const env = deps.env || stripGitHookEnv(process.env);
450
+ // The injected synchronous runner has already exited when its adapter returns,
451
+ // so there is no live child to register or reap through host process probes.
452
+ const processTree = deps.processTree || (usesSpawnSyncAdapter ? {} : createProcessTree({
453
+ env,
454
+ platform: deps.platform,
455
+ }));
456
+ let signal = null;
457
+ const removeSignalHandlers = typeof processTree.installSignalHandlers === 'function'
458
+ ? processTree.installSignalHandlers((received) => {
459
+ signal = received;
460
+ })
461
+ : () => {};
462
+ const childEnv = typeof processTree.envFor === 'function' ? processTree.envFor(env) : env;
358
463
  const bunCommand = deps.bunCommand || env.BUN_EXE || process.env.BUN_EXE || 'bun';
359
464
  const label = deps.label || 'tests';
360
465
  const timeout = resolveCommandTimeoutMs(env);
361
- const laneOptions = { env, killSignal: 'SIGKILL', timeout };
466
+ const laneOptions = { env: childEnv, killSignal: 'SIGKILL', timeout, processTree };
362
467
  // The full-suite fallback gets a larger, validation-aligned budget so a
363
468
  // healthy-but-slow full run is not failed fast by the targeted-lane ceiling.
364
- const fullSuiteOptions = { env, killSignal: 'SIGKILL', timeout: resolveFullSuiteTimeoutMs(env) };
469
+ const fullSuiteOptions = {
470
+ env: childEnv,
471
+ killSignal: 'SIGKILL',
472
+ timeout: resolveFullSuiteTimeoutMs(env),
473
+ processTree,
474
+ };
365
475
 
366
476
  console.log(`Running ${label} (${pkgManager})...`);
367
477
 
368
478
  try {
369
479
  if (plan.runFullSuite) {
370
480
  console.log(` Mode: full suite (${plan.reason})`);
371
- const status = runCommand('node', ['scripts/test-full-suite.js'], fullSuiteOptions, spawnSync);
481
+ const status = await runCommand('node', ['scripts/test-full-suite.js'], fullSuiteOptions, spawn);
482
+ if (signal) return signalExitCode(signal);
372
483
  if (status !== 0) return status;
373
484
  } else if (plan.testTargets.length > 0) {
374
485
  console.log(` Mode: targeted (${plan.testTargets.length} test file${plan.testTargets.length === 1 ? '' : 's'})`);
375
486
  const command = pkgManager === 'bun' ? bunCommand : pkgManager;
376
- const status = runCommand(command, ['run', 'test', ...plan.testTargets], laneOptions, spawnSync);
487
+ const status = await runCommand(command, ['run', 'test', ...plan.testTargets], laneOptions, spawn);
488
+ if (signal) return signalExitCode(signal);
377
489
  if (status !== 0) return status;
378
490
  }
379
491
 
380
492
  if (plan.runE2E) {
381
493
  console.log(' Extra: running affected e2e tests');
382
- const status = runCommand(bunCommand, ['test', '--timeout', '15000', 'test/e2e/'], laneOptions, spawnSync);
494
+ const status = await runCommand(bunCommand, ['test', '--timeout', '15000', 'test/e2e/'], laneOptions, spawn);
495
+ if (signal) return signalExitCode(signal);
383
496
  if (status !== 0) return status;
384
497
  }
385
498
 
386
499
  if (!plan.runFullSuite && plan.runTestEnv) {
387
500
  console.log(' Extra: running affected edge-case tests');
388
- const status = runCommand(bunCommand, ['test', '--timeout', '15000', 'test-env/'], laneOptions, spawnSync);
501
+ const status = await runCommand(bunCommand, ['test', '--timeout', '15000', 'test-env/'], laneOptions, spawn);
502
+ if (signal) return signalExitCode(signal);
389
503
  if (status !== 0) return status;
390
504
  }
391
505
 
392
506
  console.log('Relevant tests passed');
393
- return 0;
507
+ return signal ? signalExitCode(signal) : 0;
394
508
  } catch (error) {
395
509
  console.error('');
396
510
  console.error(`Failed to run ${label}: ${error.message}`);
397
511
  console.error('');
398
- return 1;
512
+ return signal ? signalExitCode(signal) : 1;
513
+ } finally {
514
+ removeSignalHandlers();
515
+ processTree.cleanup?.(signal ? 'SIGKILL' : 'SIGTERM');
399
516
  }
400
517
  }
401
518
 
@@ -406,7 +523,15 @@ function runTestExecutionPlan(plan, deps = {}) {
406
523
  * @param {Object} [deps={}] Runtime dependencies for tests.
407
524
  * @returns {number} Exit status for pre-push tests.
408
525
  */
409
- function runPrePushTests(projectRoot = process.cwd(), deps = {}) {
526
+ async function runPrePushTests(projectRoot = process.cwd(), deps = {}) {
527
+ if (isQuickPushLane(deps.env || process.env)) {
528
+ console.log('');
529
+ console.log(' quick lane: tests skipped locally — CI runs the full matrix');
530
+ console.log(' Run `forge push` (no --quick) to test before merge.');
531
+ console.log('');
532
+ return 0;
533
+ }
534
+
410
535
  const execFileSync = deps.execFileSync || defaultExecFileSync;
411
536
  const plan = classifyPushTests(projectRoot, execFileSync);
412
537
  return runTestExecutionPlan(plan, { ...deps, label: 'pre-push tests' });
@@ -419,26 +544,34 @@ function runPrePushTests(projectRoot = process.cwd(), deps = {}) {
419
544
  * @param {Object} [deps={}] Runtime dependencies for tests.
420
545
  * @returns {number} Exit status for local validation tests.
421
546
  */
422
- function runLocalValidationTests(projectRoot = process.cwd(), deps = {}) {
547
+ async function runLocalValidationTests(projectRoot = process.cwd(), deps = {}) {
423
548
  const execFileSync = deps.execFileSync || defaultExecFileSync;
424
549
  const plan = buildTestExecutionPlan(projectRoot, execFileSync, { sinceUpstream: true });
425
550
  return runTestExecutionPlan(plan, { ...deps, label: 'local validation tests' });
426
551
  }
427
552
 
428
553
  if (require.main === module) {
429
- const exitCode = process.argv.includes('--validate')
430
- ? runLocalValidationTests()
431
- : runPrePushTests();
432
- process.exit(exitCode);
554
+ (async () => {
555
+ const exitCode = process.argv.includes('--validate')
556
+ ? await runLocalValidationTests()
557
+ : await runPrePushTests();
558
+ process.exit(exitCode);
559
+ })().catch((error) => {
560
+ console.error(error);
561
+ process.exit(1);
562
+ });
433
563
  }
434
564
 
435
565
  module.exports = {
436
566
  ALWAYS_RUN_RISK_TEST_TARGETS,
437
567
  DEFAULT_FULL_SUITE_TIMEOUT_MS,
438
568
  DEFAULT_TEST_COMMAND_TIMEOUT_MS,
569
+ QUICK_LANE_ENV_VAR,
570
+ QUICK_LANE_VALUE,
439
571
  buildTestExecutionPlan,
440
572
  classifyPushTests,
441
573
  detectPackageManager,
574
+ isQuickPushLane,
442
575
  resolveCommandTimeoutMs,
443
576
  resolveFullSuiteTimeoutMs,
444
577
  runLocalValidationTests,
@@ -1,17 +1,10 @@
1
1
  ---
2
2
  name: review
3
+ invocation: user
3
4
  description: >
4
- Forge REVIEW stage — drives an already-open PR to all-green by resolving EVERY piece of
5
- feedback: failing GitHub Actions / CI checks, Greptile and CodeRabbit inline comments and
6
- summaries, and the SonarCloud quality gate. It fixes the code, replies to each thread and
7
- marks it resolved until checks pass, then runs the pre-merge doc gate and hands the PR off
8
- for MANUAL merge — never runs `gh pr merge`. Use whenever a PR has review feedback or red
9
- checks: "address the review comments on PR #123", "the bots flagged issues / Greptile score
10
- is low", "fix the failing checks and resolve the threads", "CodeRabbit and SonarCloud left
11
- comments", "get my PR merge-ready". This MUTATING feedback-resolution stage is distinct from
12
- shepherd (only WATCHES an open PR, changes nothing), verify (POST-merge CI health check +
13
- closing issues), ship (pushes the branch + opens the PR), validate (PRE-PR local
14
- type/lint/test/security), and sonarcloud / sonarcloud-analysis (only QUERY SonarCloud data).
5
+ Drive an already-open PR's feedback and failing checks to all-green, resolve every thread,
6
+ run the pre-merge doc gate, then stop for manual merge. Use only when the user asks to
7
+ address PR feedback or checks; never merges or substitutes for Shepherd monitoring.
15
8
  allowed-tools: Bash, Read, Edit, Grep, Glob
16
9
  next: verify
17
10
  terminal: false
@@ -3,9 +3,9 @@
3
3
  "fixtures": "present",
4
4
  "static": {
5
5
  "token_cost": {
6
- "desc_chars": 988,
6
+ "desc_chars": 261,
7
7
  "body_lines": 476,
8
- "score": 4
8
+ "score": 40
9
9
  },
10
10
  "caps": {
11
11
  "desc_within": true,
@@ -37,5 +37,5 @@
37
37
  "variance": null,
38
38
  "note": "behavioral — W5 (LLM judge): semantic recall/precision/chain/outcome/variance, not deterministic"
39
39
  },
40
- "composite": 71
40
+ "composite": 82
41
41
  }
@@ -1,17 +1,10 @@
1
1
  ---
2
2
  name: rollback
3
+ invocation: user
3
4
  description: >
4
- Safely revert or undo a change that is already committed or shipped, using non-destructive
5
- `git revert` (never `reset --hard`, `push --force`, or `clean`). Drives the interactive
6
- `bunx forge rollback` menu: undo the last commit, revert a commit by hash, revert a merged
7
- PR (git revert -m 1), restore individual files to a prior version, revert a commit range, or
8
- dry-run/preview first. Use when the user says roll back, revert, undo, back out, or restore
9
- — e.g. "undo the last commit", "undo that change, the approach was wrong", "back out PR
10
- #123", "revert the merge commit", "restore src/auth.js to before my last commit", or when a
11
- shipped/merged change broke something and must be pulled back. Do NOT use for: replying to
12
- or fixing PR review feedback (review), monitoring an open PR's checks toward merge
13
- (shepherd), the post-merge CI health check on master (verify), pushing a branch or opening a
14
- PR (ship), or ordinary issue status/field edits (issue-basics).
5
+ Safely undo committed or shipped changes with non-destructive git revert or restore. Use
6
+ only when the user explicitly asks to roll back, revert, undo, back out, or restore; never
7
+ reset hard, force-push, or substitute for PR review, monitoring, or issue status work.
15
8
  allowed-tools: Bash, Read, Edit, Grep, Glob
16
9
  terminal: true
17
10
  ---
@@ -3,9 +3,9 @@
3
3
  "fixtures": "present",
4
4
  "static": {
5
5
  "token_cost": {
6
- "desc_chars": 965,
6
+ "desc_chars": 266,
7
7
  "body_lines": 95,
8
- "score": 43
8
+ "score": 78
9
9
  },
10
10
  "caps": {
11
11
  "desc_within": true,
@@ -37,5 +37,5 @@
37
37
  "variance": null,
38
38
  "note": "behavioral — W5 (LLM judge): semantic recall/precision/chain/outcome/variance, not deterministic"
39
39
  },
40
- "composite": 83
40
+ "composite": 93
41
41
  }
@@ -6,8 +6,8 @@ description: >
6
6
  open PR, converges CI check state into kernel verdicts, re-runs flaky required
7
7
  checks, reaps orphan watchers, and self-retires when no PRs remain; one-shot
8
8
  passes (forge shepherd <pr>) exist for a single bounded check. Use when a PR was
9
- just opened or shipped; when a session starts with open PRs (start the daemon in
10
- a background shell — the singleton lease makes a duplicate start a safe no-op);
9
+ just opened or shipped; when a session starts with open PRs (Forge automatically
10
+ wakes the singleton; the lease makes a duplicate start a safe no-op);
11
11
  when the user asks "why isn't my PR merging", "what's
12
12
  blocking the PR", "is the PR ready", "did a check fail / go red"; when asked to
13
13
  "keep watching / keep an eye on / babysit my PRs"; or to read PR verdicts
@@ -34,15 +34,18 @@ Forge owns your open PRs. A singleton daemon converges every open PR toward merg
34
34
  - User asks about **one PR, right now** ("is #212 ready?", "why is this blocked?") → **one-shot** (`forge shepherd <pr> --pull`).
35
35
  - Anything **ongoing** — "keep", "until it merges", "watch", "babysit", multiple PRs, or a session that starts with PRs already open → **ensure the daemon is running**.
36
36
 
37
- ## Session-start ritual
37
+ ## Automatic attachment
38
38
 
39
- When a session starts and the repo has open PRs, ensure the daemon is running:
39
+ Forge wakes the singleton after a successful supported session start, every
40
+ successful push, and every successful non-dry-run ship. Push does not require a
41
+ locally resolved PR number; the daemon enumerates all open and draft PRs itself.
42
+ Ordinary commands do not launch it.
40
43
 
41
- ```bash
42
- forge shepherd daemon # start it in the HARNESS BACKGROUND SHELL
43
- ```
44
-
45
- The daemon is a **repo singleton** guarded by an O_EXCL lease, so you do NOT need a liveness check first — a second start when one is already running simply exits (the lease is already held). Start it in the harness background shell (Claude Code / Codex background-shell) so it is session-scoped and reaped with the session. NEVER launch a detached spawn from the agent: the detached path is Forge's bare-CLI fallback only. (Automatic per-command launch and a `forge prime` daemon-liveness line are planned follow-ups — W-S4c/W-S5 — not yet wired; until then you start the daemon explicitly as above.)
44
+ The daemon is a **repo singleton** guarded by an O_EXCL lease, so no liveness
45
+ check is needed first. Forge prefers a supplied harness background shell; its
46
+ bare-CLI detached fallback launches from the stable common repository root, not
47
+ the current disposable worktree. Agents never launch a detached spawn directly.
48
+ (A `forge prime` daemon-liveness line remains the separate W-S5 follow-up.)
46
49
 
47
50
  ## Reading verdicts (the common case)
48
51
 
@@ -69,8 +72,8 @@ forge shepherd events <pr> --since <seq> # only the new events since sequence
69
72
 
70
73
  | Situation | Command |
71
74
  | --- | --- |
72
- | PR just opened / shipped | ensure `forge shepherd daemon` running |
73
- | Session starts, open PRs exist, daemon dead | `forge shepherd daemon` (background shell) |
75
+ | PR just opened / shipped | automatic singleton wake; read with `forge shepherd <pr> --pull` |
76
+ | Session starts with open PRs | automatic singleton wake |
74
77
  | "Why isn't my PR merging / what's blocking it" | `forge shepherd <pr> --pull` |
75
78
  | "Is the PR ready?" | `forge shepherd <pr> --pull` (read `MERGE_READY`) |
76
79
  | "A check failed / went red" | `forge shepherd <pr> --pull --json` (read `failures[]`) |
@@ -95,11 +98,14 @@ forge shepherd events <pr> --since <seq> # only the new events since sequence
95
98
  ## Kill-switches
96
99
 
97
100
  ```bash
98
- FORGE_SHEPHERD_DISABLE=1 # env: makes the shepherd trigger inert (once the auto-fire wiring lands, W-S4c)
99
- forge gate disable rail.auto_shepherd # config gate honored by the trigger + ship/push arming
101
+ FORGE_SHEPHERD_DISABLE=1 # env: makes automatic firing inert
102
+ forge gate disable rail.auto_shepherd # config gate honored by every automatic seam
100
103
  ```
101
104
 
102
- Both leave the manual `forge shepherd` surface usable; they only stop the automatic daemon fire.
105
+ CI/test (`NODE_ENV=test`, `BUN_ENV=test`, `CI`, `GITHUB_ACTIONS`, or
106
+ `GITLAB_CI`), dry-run, uninitialized repositories, and disabled paths return
107
+ before lease, Kernel-state, or process work. All leave the manual
108
+ `forge shepherd` surface usable; they only stop automatic daemon fire.
103
109
 
104
110
  ## State
105
111
 
@@ -3,8 +3,8 @@
3
3
  "fixtures": "present",
4
4
  "static": {
5
5
  "token_cost": {
6
- "desc_chars": 997,
7
- "body_lines": 86,
6
+ "desc_chars": 987,
7
+ "body_lines": 92,
8
8
  "score": 43
9
9
  },
10
10
  "caps": {
@@ -1,18 +1,10 @@
1
1
  ---
2
2
  name: ship
3
+ invocation: user
3
4
  description: >
4
- Forge SHIP stage: push the validated feature branch and open a PR populated from the
5
- project's OWN PR template (design-doc link, Forge issue IDs, real test/commit data), then
6
- hand off for MANUAL merge; never merges or auto-merges. Use once /validate passes and you
7
- want a PR on the board. Triggers: "ship it", "ship this branch", "open the PR", "push and
8
- open a PR", "gh pr create", "checks passed, now cut the PR". Runs branch-freshness +
9
- parallel-PR merge-sim checks, force-with-lease push, records the ship->review handoff, then
10
- stops. One stage only; not for: the whole plan->dev->validate->ship->review pipeline or
11
- drive-to-done (smith); type-check/lint/tests/security first (validate); addressing PR
12
- comments or resolving Greptile/SonarCloud/CodeRabbit threads on an existing PR (review);
13
- babysitting an open PR toward merge (shepherd); post-merge CI health check + closing issues
14
- (verify); reverting an already-shipped change (rollback). If the PR already exists, this is
15
- not the skill.
5
+ Push a validated feature branch and open its PR from the project's template, then stop for
6
+ manual merge. Use only when the user asks to ship or open the PR after validation; never
7
+ merges, auto-merges, reviews an existing PR, or replaces end-to-end Smith.
16
8
  allowed-tools: Bash, Read, Edit, Grep, Glob
17
9
  next: review
18
10
  terminal: false
@@ -3,9 +3,9 @@
3
3
  "fixtures": "present",
4
4
  "static": {
5
5
  "token_cost": {
6
- "desc_chars": 993,
6
+ "desc_chars": 254,
7
7
  "body_lines": 237,
8
- "score": 28
8
+ "score": 64
9
9
  },
10
10
  "caps": {
11
11
  "desc_within": true,
@@ -37,5 +37,5 @@
37
37
  "variance": null,
38
38
  "note": "behavioral — W5 (LLM judge): semantic recall/precision/chain/outcome/variance, not deterministic"
39
39
  },
40
- "composite": 78
40
+ "composite": 89
41
41
  }
@@ -71,7 +71,12 @@ forge clean # remove worktrees whose branches merged; FF th
71
71
  forge clean --dry-run # show what would be removed, change nothing
72
72
  ```
73
73
 
74
- `forge clean` is **squash-aware** — it detects branches that merged via squash, not just fast-forward merges.
74
+ `forge clean` is **squash-aware** — it detects branches that merged via squash, not just fast-forward merges. Removing a merged worktree also closes the kernel issue the branch was linked to, citing the merge as evidence.
75
+
76
+ Two things it will never touch, because closing an issue is irreversible (`done` is terminal):
77
+
78
+ - **A branch with no commits of its own.** A freshly created worktree points at a default-branch commit, which git alone reads as "merged". That is unstarted work — it is kept, and its issue stays open.
79
+ - **A worktree it skipped.** Uncommitted changes hold a worktree back, and a held-back worktree never has its issue closed.
75
80
 
76
81
  ### `forge clean` footguns
77
82
 
@@ -4,8 +4,8 @@
4
4
  "static": {
5
5
  "token_cost": {
6
6
  "desc_chars": 1013,
7
- "body_lines": 70,
8
- "score": 44
7
+ "body_lines": 75,
8
+ "score": 43
9
9
  },
10
10
  "caps": {
11
11
  "desc_within": true,