mandrel 2.111.0 → 2.113.0

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 (183) hide show
  1. package/.agents/agents/acceptance-critic.md +14 -0
  2. package/.agents/agents/auditor.md +14 -0
  3. package/.agents/agents/plan-critic.md +5 -0
  4. package/.agents/agents/story-worker.md +4 -1
  5. package/.agents/docs/agentrc-reference.json +22 -10
  6. package/.agents/docs/configuration.md +367 -37
  7. package/.agents/docs/execution-reference.md +11 -0
  8. package/.agents/rules/git-conventions-reference.md +7 -0
  9. package/.agents/schemas/agentrc.schema.json +174 -14
  10. package/.agents/schemas/baselines/baseline-envelope.schema.json +16 -4
  11. package/.agents/schemas/baselines/bundle-size.schema.json +1 -0
  12. package/.agents/schemas/baselines/coverage.schema.json +1 -0
  13. package/.agents/schemas/baselines/crap.schema.json +37 -10
  14. package/.agents/schemas/baselines/duplication.schema.json +1 -0
  15. package/.agents/schemas/baselines/maintainability.schema.json +1 -0
  16. package/.agents/schemas/baselines/mutation.schema.json +1 -0
  17. package/.agents/schemas/crap-report.schema.json +6 -5
  18. package/.agents/schemas/story-deliver-terminal.schema.json +41 -2
  19. package/.agents/schemas/story-progress.schema.json +1 -1
  20. package/.agents/schemas/validation-evidence.schema.json +2 -2
  21. package/.agents/scripts/boot-sweep.js +41 -3
  22. package/.agents/scripts/coverage-capture.js +37 -3
  23. package/.agents/scripts/coverage-ratchet.js +106 -0
  24. package/.agents/scripts/evidence-gate.js +84 -21
  25. package/.agents/scripts/lib/audit-baselines/read.js +8 -1
  26. package/.agents/scripts/lib/audit-baselines/trend.js +2 -1
  27. package/.agents/scripts/lib/baseline-loader.js +4 -2
  28. package/.agents/scripts/lib/baselines/codec.js +57 -0
  29. package/.agents/scripts/lib/baselines/crap-preview-incremental.js +8 -3
  30. package/.agents/scripts/lib/baselines/crap-preview-scan.js +10 -16
  31. package/.agents/scripts/lib/baselines/crap-updater-cli.js +4 -13
  32. package/.agents/scripts/lib/baselines/kernel.js +19 -1
  33. package/.agents/scripts/lib/baselines/kinds/_crap-codec.js +171 -0
  34. package/.agents/scripts/lib/baselines/kinds/_crap-read.js +56 -16
  35. package/.agents/scripts/lib/baselines/kinds/_shared-metric.js +0 -59
  36. package/.agents/scripts/lib/baselines/kinds/crap.js +283 -319
  37. package/.agents/scripts/lib/baselines/merge-envelopes.js +46 -1
  38. package/.agents/scripts/lib/baselines/preview-gates.js +29 -10
  39. package/.agents/scripts/lib/baselines/reader.js +36 -9
  40. package/.agents/scripts/lib/baselines/refresh-service.js +93 -21
  41. package/.agents/scripts/lib/baselines/seat-missing.js +173 -7
  42. package/.agents/scripts/lib/baselines/writer.js +36 -21
  43. package/.agents/scripts/lib/changed-files.js +70 -6
  44. package/.agents/scripts/lib/child-exec.js +10 -0
  45. package/.agents/scripts/lib/claude/agents-projection.js +5 -5
  46. package/.agents/scripts/lib/close-lock-events.js +8 -3
  47. package/.agents/scripts/lib/close-validation/ci-deferral.js +40 -0
  48. package/.agents/scripts/lib/close-validation/gates.js +111 -35
  49. package/.agents/scripts/lib/close-validation/process.js +16 -7
  50. package/.agents/scripts/lib/close-validation/projections/crap.js +43 -10
  51. package/.agents/scripts/lib/close-validation/runner.js +55 -23
  52. package/.agents/scripts/lib/close-validation/skip-branch-test.js +144 -0
  53. package/.agents/scripts/lib/close-validation/test-delta.js +33 -8
  54. package/.agents/scripts/lib/codex/agents-projection.js +6 -0
  55. package/.agents/scripts/lib/config/agentrc-overlay.js +9 -5
  56. package/.agents/scripts/lib/config/explain.js +5 -2
  57. package/.agents/scripts/lib/config/gates/coverage.schema.js +45 -1
  58. package/.agents/scripts/lib/config/gates/crap-incremental-coverage.schema.js +1 -1
  59. package/.agents/scripts/lib/config/git-roots.js +38 -4
  60. package/.agents/scripts/lib/config/handoff.js +24 -0
  61. package/.agents/scripts/lib/config/quality.js +26 -13
  62. package/.agents/scripts/lib/config/worktree-isolation.js +2 -5
  63. package/.agents/scripts/lib/config-resolver.js +22 -2
  64. package/.agents/scripts/lib/config-settings-schema-delivery.js +87 -4
  65. package/.agents/scripts/lib/config-settings-schema-quality.js +36 -1
  66. package/.agents/scripts/lib/config-settings-schema.js +1 -1
  67. package/.agents/scripts/lib/coverage-capture-affected.js +1 -0
  68. package/.agents/scripts/lib/coverage-capture-delta.js +164 -32
  69. package/.agents/scripts/lib/coverage-capture-own-fix.js +232 -0
  70. package/.agents/scripts/lib/coverage-capture-usage.js +2 -2
  71. package/.agents/scripts/lib/coverage-capture.js +15 -0
  72. package/.agents/scripts/lib/coverage-deferral.js +96 -0
  73. package/.agents/scripts/lib/coverage-ratchet.js +187 -0
  74. package/.agents/scripts/lib/crap-baseline-join.js +5 -5
  75. package/.agents/scripts/lib/crap-engine.js +3 -3
  76. package/.agents/scripts/lib/crap-method-identity.js +104 -48
  77. package/.agents/scripts/lib/error-redactor.js +20 -0
  78. package/.agents/scripts/lib/escomplex-anchor.js +135 -0
  79. package/.agents/scripts/lib/escomplex-kernel.js +30 -1
  80. package/.agents/scripts/lib/failure-rows.js +121 -0
  81. package/.agents/scripts/lib/full-suite-lock.js +44 -4
  82. package/.agents/scripts/lib/full-suite-queue.js +89 -13
  83. package/.agents/scripts/lib/generated/agentrc-validator.js +8 -2
  84. package/.agents/scripts/lib/gh-env.js +3 -6
  85. package/.agents/scripts/lib/gh-pr-rest.js +1 -0
  86. package/.agents/scripts/lib/git/base-sync-probe.js +182 -0
  87. package/.agents/scripts/lib/git/baseline-conflict-rows.js +60 -0
  88. package/.agents/scripts/lib/git/sync-from-base.js +18 -9
  89. package/.agents/scripts/lib/git-branch-lifecycle.js +9 -0
  90. package/.agents/scripts/lib/observability/close-telemetry.js +37 -2
  91. package/.agents/scripts/lib/observability/handoff-telemetry.js +170 -0
  92. package/.agents/scripts/lib/observability/review-dispositions.js +23 -6
  93. package/.agents/scripts/lib/observability/runtime-friction.js +17 -4
  94. package/.agents/scripts/lib/observability/source-classifier.js +3 -0
  95. package/.agents/scripts/lib/observability/story-outcome.js +393 -0
  96. package/.agents/scripts/lib/orchestration/check-baselines/phases/compare.js +2 -1
  97. package/.agents/scripts/lib/orchestration/check-baselines/phases/evaluate.js +19 -1
  98. package/.agents/scripts/lib/orchestration/check-baselines/phases/fail-detail.js +84 -0
  99. package/.agents/scripts/lib/orchestration/check-baselines/phases/pipeline.js +14 -5
  100. package/.agents/scripts/lib/orchestration/check-baselines/phases/refresh-ack.js +3 -2
  101. package/.agents/scripts/lib/orchestration/check-baselines/phases/report.js +26 -6
  102. package/.agents/scripts/lib/orchestration/code-review.js +223 -19
  103. package/.agents/scripts/lib/orchestration/handoff-gates.js +129 -62
  104. package/.agents/scripts/lib/orchestration/handoff-rescore.js +170 -0
  105. package/.agents/scripts/lib/orchestration/handoff-suite-of-record.js +45 -0
  106. package/.agents/scripts/lib/orchestration/preflight-lanes.js +208 -0
  107. package/.agents/scripts/lib/orchestration/review-base-ref.js +40 -17
  108. package/.agents/scripts/lib/orchestration/review-deposit.js +38 -12
  109. package/.agents/scripts/lib/orchestration/review-depth.js +21 -0
  110. package/.agents/scripts/lib/orchestration/review-history.js +124 -20
  111. package/.agents/scripts/lib/orchestration/review-providers/claude-verdict.js +41 -12
  112. package/.agents/scripts/lib/orchestration/review-providers/code-review.js +90 -39
  113. package/.agents/scripts/lib/orchestration/review-providers/critical-verifier.js +64 -11
  114. package/.agents/scripts/lib/orchestration/review-providers/degraded-gates.js +102 -3
  115. package/.agents/scripts/lib/orchestration/review-providers/finding-identity.js +6 -2
  116. package/.agents/scripts/lib/orchestration/review-providers/findings-renderer.js +55 -8
  117. package/.agents/scripts/lib/orchestration/review-providers/review-guidance.js +1 -0
  118. package/.agents/scripts/lib/orchestration/review-providers/review-provider-factory.js +42 -15
  119. package/.agents/scripts/lib/orchestration/review-providers/types.js +6 -0
  120. package/.agents/scripts/lib/orchestration/single-story-close/failed-terminal.js +10 -1
  121. package/.agents/scripts/lib/orchestration/single-story-close/gate-log.js +10 -0
  122. package/.agents/scripts/lib/orchestration/single-story-close/phases/base-sync.js +7 -4
  123. package/.agents/scripts/lib/orchestration/single-story-close/phases/close-validation.js +16 -3
  124. package/.agents/scripts/lib/orchestration/single-story-close/phases/code-review.js +59 -2
  125. package/.agents/scripts/lib/orchestration/single-story-close/phases/confirm-merge.js +6 -0
  126. package/.agents/scripts/lib/orchestration/single-story-close/phases/finish.js +15 -4
  127. package/.agents/scripts/lib/orchestration/single-story-close/phases/land-review-record.js +213 -0
  128. package/.agents/scripts/lib/orchestration/single-story-close/phases/normalize-pr-title.js +9 -12
  129. package/.agents/scripts/lib/orchestration/single-story-close/phases/open-pr.js +26 -12
  130. package/.agents/scripts/lib/orchestration/single-story-close/phases/post-land.js +51 -26
  131. package/.agents/scripts/lib/orchestration/single-story-close/phases/pre-gate-steps.js +2 -1
  132. package/.agents/scripts/lib/orchestration/single-story-close/phases/pre-push.js +3 -2
  133. package/.agents/scripts/lib/orchestration/single-story-close/phases/pull-request.js +50 -17
  134. package/.agents/scripts/lib/orchestration/single-story-close/phases/push.js +59 -11
  135. package/.agents/scripts/lib/orchestration/single-story-close/phases/reconcile-done.js +1 -0
  136. package/.agents/scripts/lib/orchestration/single-story-close/phases/review-block.js +31 -13
  137. package/.agents/scripts/lib/orchestration/single-story-close/phases/review-halt.js +29 -6
  138. package/.agents/scripts/lib/orchestration/single-story-close/phases/review-outcome.js +10 -12
  139. package/.agents/scripts/lib/orchestration/single-story-close/phases/review-override.js +19 -4
  140. package/.agents/scripts/lib/orchestration/single-story-close/result.js +30 -13
  141. package/.agents/scripts/lib/orchestration/single-story-close/review-overlap.js +73 -39
  142. package/.agents/scripts/lib/orchestration/single-story-close/runner.js +11 -3
  143. package/.agents/scripts/lib/orchestration/story-close/baseline-upward-writeback.js +3 -2
  144. package/.agents/scripts/lib/orchestration/story-close/duplication-close-rescore.js +77 -15
  145. package/.agents/scripts/lib/orchestration/story-deliver-terminal.js +7 -0
  146. package/.agents/scripts/lib/orchestration/story-handoff.js +796 -196
  147. package/.agents/scripts/lib/orchestration/ticketing/reads.js +2 -0
  148. package/.agents/scripts/lib/pr-body-lines.js +32 -0
  149. package/.agents/scripts/lib/single-story/ci-ratchet-gate.js +140 -0
  150. package/.agents/scripts/lib/single-story/confirm-and-land.js +17 -7
  151. package/.agents/scripts/lib/single-story/confirm-merge.js +12 -3
  152. package/.agents/scripts/lib/single-story-sweep/report-stalled-executing.js +202 -0
  153. package/.agents/scripts/lib/standalone-ratchets.js +37 -0
  154. package/.agents/scripts/lib/suite-admission.js +20 -8
  155. package/.agents/scripts/lib/suite-sizing.js +146 -0
  156. package/.agents/scripts/lib/workspace-provisioner.js +2 -2
  157. package/.agents/scripts/merge-baseline.js +14 -3
  158. package/.agents/scripts/providers/github/comments.js +19 -3
  159. package/.agents/scripts/providers/github.js +1 -0
  160. package/.agents/scripts/review-dispositions-report.js +1 -1
  161. package/.agents/scripts/role-scope-guard.js +267 -0
  162. package/.agents/scripts/single-story-close.js +4 -4
  163. package/.agents/scripts/single-story-confirm-merge.js +31 -5
  164. package/.agents/scripts/single-story-init.js +24 -4
  165. package/.agents/scripts/story-handoff.js +3 -3
  166. package/.agents/scripts/story-outcomes-report.js +231 -0
  167. package/.agents/scripts/story-review-compute.js +9 -5
  168. package/.agents/scripts/update-coverage-baseline.js +12 -2
  169. package/.agents/scripts/update-crap-baseline.js +9 -1
  170. package/.agents/workflows/helpers/code-review.md +63 -9
  171. package/.agents/workflows/helpers/deliver-digest.md +53 -54
  172. package/.agents/workflows/helpers/deliver-reference.md +233 -67
  173. package/.agents/workflows/helpers/deliver-story-reference.md +19 -8
  174. package/.agents/workflows/helpers/deliver-story.md +7 -2
  175. package/.agents/workflows/helpers/parallel-tooling.md +29 -0
  176. package/docs/CHANGELOG.md +55 -0
  177. package/lib/cli/doctor-checks/coverage-ratchet.js +54 -0
  178. package/lib/cli/doctor-checks/effective-agentrc.js +23 -0
  179. package/lib/cli/doctor-checks/llm-review.js +2 -11
  180. package/lib/cli/doctor-checks/sync-parity.js +5 -1
  181. package/lib/cli/doctor-checks/test-credit-path.js +7 -2
  182. package/lib/cli/registry.js +5 -0
  183. package/package.json +1 -1
@@ -5,6 +5,7 @@
5
5
  * @typedef {import('./types.js').Finding} Finding
6
6
  */
7
7
 
8
+ import { redactPii } from '../../error-redactor.js';
8
9
  import { isObject } from '../../json-utils.js';
9
10
  import { CLAIM_SCHEMA } from './claim-grounding.js';
10
11
  import { buildFinding, unwrapEnvelope } from './parse-findings.js';
@@ -63,13 +64,21 @@ function tryParseJson(text) {
63
64
  }
64
65
  }
65
66
 
67
+ const AUTH_PREFIX_CHARS = 200;
68
+ const AUTH_RETRY_DELAY_MS = 2000;
69
+
70
+ /** A PII-free, length-capped prefix of raw CLI output. */
71
+ const prefixOf = (text) =>
72
+ redactPii(text).replace(/\s+/g, ' ').trim().slice(0, AUTH_PREFIX_CHARS);
73
+
66
74
  /**
67
75
  * `claude auth status --json`. Only `authMethod` and `subscriptionType`
68
76
  * are ever named: the payload also carries the account email, org name
69
- * and ids, which no reason or log line may repeat.
77
+ * and ids, which no reason or log line may repeat. Non-JSON output on a
78
+ * zero exit is transient, so it is `retryable` after `delayMs`.
70
79
  *
71
- * @param {{ status: number, stdout: string }} result
72
- * @returns {{ ok: true, detail: string } | { ok: false, reason: string }}
80
+ * @param {{ status: number, stdout: string, stderr?: string }} result
81
+ * @returns {{ ok: true, detail: string } | { ok: false, reason: string, retryable?: boolean, delayMs?: number }}
73
82
  */
74
83
  export function readAuthStatus(result) {
75
84
  if (result.status !== 0) {
@@ -82,7 +91,12 @@ export function readAuthStatus(result) {
82
91
  if (!parsed || typeof parsed !== 'object') {
83
92
  return {
84
93
  ok: false,
85
- reason: 'claude auth status returned unparseable output',
94
+ reason:
95
+ 'claude auth status returned unparseable output ' +
96
+ `(stdout: "${prefixOf(result.stdout)}", ` +
97
+ `stderr: "${prefixOf(result.stderr)}")`,
98
+ retryable: true,
99
+ delayMs: AUTH_RETRY_DELAY_MS,
86
100
  };
87
101
  }
88
102
  const method =
@@ -187,11 +201,21 @@ function toFindings(entries) {
187
201
  .filter(Boolean);
188
202
  }
189
203
 
204
+ const resultTextOf = (envelope) =>
205
+ typeof envelope?.result === 'string' ? envelope.result : '';
206
+
207
+ function isUsageLimited(result, envelope) {
208
+ const text = [result.stderr, resultTextOf(envelope), result.stdout];
209
+ return USAGE_LIMIT_RE.test(text.join('\n'));
210
+ }
211
+
212
+ function isRetryable(result, envelope) {
213
+ return Number.isInteger(result.status) && !isUsageLimited(result, envelope);
214
+ }
215
+
190
216
  function failureReason(result, envelope) {
191
- const resultText =
192
- typeof envelope?.result === 'string' ? envelope.result : '';
193
- const text = [result.stderr, resultText, result.stdout].join('\n');
194
- if (USAGE_LIMIT_RE.test(text)) {
217
+ const resultText = resultTextOf(envelope);
218
+ if (isUsageLimited(result, envelope)) {
195
219
  return 'the claude CLI account is out of quota (usage limit reached)';
196
220
  }
197
221
  if (result.status !== 0) {
@@ -237,19 +261,24 @@ function resultTextFindings(text) {
237
261
  /**
238
262
  * Verdict resolution: (a) a successful envelope's `structured_output`,
239
263
  * (b) the fallback parser over its `result` text, (c) no verdict.
264
+ * `retryable` excludes a usage limit and a `status: null` timeout.
240
265
  *
241
- * @param {{ status: number, stdout: string, stderr: string }} result
242
- * @returns {{ findings: Finding[] } | { reason: string }}
266
+ * @param {{ status: number|null, stdout: string, stderr: string }} result
267
+ * @returns {{ findings: Finding[] } | { reason: string, retryable: boolean }}
243
268
  */
244
269
  export function resolveVerdict(result) {
245
270
  const envelope = parseEnvelope(result.stdout);
271
+ const retryable = isRetryable(result, envelope);
246
272
  if (!isSuccess(result, envelope)) {
247
- return { reason: noVerdictReason(result, envelope) };
273
+ return { reason: noVerdictReason(result, envelope), retryable };
248
274
  }
249
275
  const entries =
250
276
  structuredFindings(envelope.structured_output) ??
251
277
  resultTextFindings(envelope.result);
252
278
  return entries
253
279
  ? { findings: toFindings(entries) }
254
- : { reason: 'the reviewer output did not parse as a findings verdict' };
280
+ : {
281
+ reason: 'the reviewer output did not parse as a findings verdict',
282
+ retryable,
283
+ };
255
284
  }
@@ -1,6 +1,6 @@
1
1
  /**
2
- * review-providers/code-review.js — a low-effort model bug review of the
3
- * Story diff through `claude --print --effort low`. It asks only for
2
+ * review-providers/code-review.js — a model bug review of the Story diff
3
+ * through `claude --print --effort <input.effort ?? low>`. It asks only for
4
4
  * merge-blocking problems, so every parsed finding is `critical` and halts
5
5
  * close before auto-merge. The reviewer is handed the diff range, the Story
6
6
  * id and the diff text — never `acceptance[]` or the self-eval verdict: this
@@ -10,14 +10,15 @@
10
10
  * OS temp cwd (no CLAUDE.md discovery), user-only settings, no MCP, no tools.
11
11
  *
12
12
  * A missing CLI throws at construction (the chain entry is optional). A
13
- * review with no verdict records one degraded gate via `getDegradations()`
14
- * and returns no findings, so "never ran" cannot read as "found nothing".
13
+ * transient failure is retried once; a review still with no verdict records
14
+ * one degraded gate via `getDegradations()` and returns no findings.
15
15
  * A diff over the cap is reviewed chunk by chunk (file-aligned, sequential,
16
16
  * findings merged). Source sections pack before tests and docs, and every
17
17
  * chunk is told it is a slice and handed the whole changed-file manifest so
18
- * a file changed in another chunk never reads as unchanged. What went unchecked — a truncated file, files past the
19
- * chunk bound, a chunk with no verdict, a capped `<callers>` block — is
20
- * recorded beside the findings, naming the files.
18
+ * a file changed in another chunk never reads as unchanged. What went
19
+ * unchecked — a truncated file, files past the chunk bound, a chunk with no
20
+ * verdict (`coverage-partial`), a capped `<callers>` block
21
+ * (`context-reduced`) — is recorded beside the findings, naming the files.
21
22
  *
22
23
  * @typedef {import('./types.js').Finding} Finding
23
24
  * @typedef {import('./types.js').ReviewInput} ReviewInput
@@ -28,9 +29,13 @@
28
29
  import fs from 'node:fs';
29
30
  import os from 'node:os';
30
31
  import path from 'node:path';
31
- import { spawnCaptureAsync } from '../../child-exec.js';
32
+ import {
33
+ spawnCaptureAsync,
34
+ spawnCaptureNullableAsync,
35
+ } from '../../child-exec.js';
32
36
  import { gitSpawn } from '../../git-utils.js';
33
37
  import { PROJECT_ROOT } from '../../project-root.js';
38
+ import { defaultSleep } from '../../util/sleep.js';
34
39
  import { buildCallersSection } from './callers-context.js';
35
40
  import {
36
41
  FINDINGS_SCHEMA_JSON,
@@ -43,10 +48,11 @@ import { renderDepthDirective } from './review-depth.js';
43
48
  import { renderGuidanceBlock } from './review-guidance.js';
44
49
  import { probeClaudeCli } from './security-review.js';
45
50
 
46
- const CLAUDE_ARGS = Object.freeze([
51
+ /** @param {string} effort */
52
+ const claudeArgs = (effort) => [
47
53
  '--print',
48
54
  '--effort',
49
- 'low',
55
+ effort,
50
56
  '--output-format',
51
57
  'json',
52
58
  '--json-schema',
@@ -56,7 +62,7 @@ const CLAUDE_ARGS = Object.freeze([
56
62
  '--strict-mcp-config',
57
63
  '--tools',
58
64
  '',
59
- ]);
65
+ ];
60
66
  const AUTH_ARGS = Object.freeze(['auth', 'status', '--json']);
61
67
  const INVOKE_TIMEOUT_MS = 10 * 60 * 1000;
62
68
  const AUTH_TIMEOUT_MS = 15 * 1000;
@@ -239,19 +245,22 @@ function quoteWindowsArg(arg) {
239
245
  * @param {{ cwd: string, input?: string, timeout: number, run?: Function, platform: string }} opts
240
246
  * @returns {Promise<{ status: number, stdout: string, stderr: string }>}
241
247
  */
242
- export function runClaude(args, { cwd, input, timeout, run, platform }) {
248
+ export function runClaude(args, opts) {
249
+ return spawnCaptureAsync('claude', ...claudeSpawn(args, opts));
250
+ }
251
+
252
+ /** `runClaude`, but a timeout, kill or spawn error stays `status: null`. */
253
+ function runClaudeNullable(args, opts) {
254
+ return spawnCaptureNullableAsync('claude', ...claudeSpawn(args, opts));
255
+ }
256
+
257
+ /** The argv and spawn options both runners share. */
258
+ function claudeSpawn(args, { cwd, input, timeout, run, platform }) {
243
259
  const shell = platform === 'win32';
244
- return spawnCaptureAsync(
245
- 'claude',
260
+ return [
246
261
  shell ? args.map(quoteWindowsArg) : [...args],
247
- {
248
- cwd,
249
- input: input ?? '',
250
- shell,
251
- timeout,
252
- ...(run ? { run } : {}),
253
- },
254
- );
262
+ { cwd, input: input ?? '', shell, timeout, ...(run ? { run } : {}) },
263
+ ];
255
264
  }
256
265
 
257
266
  /**
@@ -280,8 +289,8 @@ function runAuthStatus(runDir, deps) {
280
289
  });
281
290
  }
282
291
 
283
- function runReviewer(runDir, prompt, deps) {
284
- return runClaude(CLAUDE_ARGS, {
292
+ function runReviewer(runDir, prompt, effort, deps) {
293
+ return runClaudeNullable(claudeArgs(effort), {
285
294
  cwd: runDir,
286
295
  input: prompt,
287
296
  timeout: INVOKE_TIMEOUT_MS,
@@ -293,22 +302,41 @@ function runReviewer(runDir, prompt, deps) {
293
302
  /**
294
303
  * Preflight, then the review, both from a fresh OS temp dir removed after.
295
304
  *
296
- * @returns {Promise<{ findings: Finding[] } | { reason: string }>}
305
+ * @returns {Promise<{ findings: Finding[] } | { reason: string, retryable?: boolean }>}
297
306
  */
298
- async function reviewInTempDir(prompt, deps) {
307
+ async function reviewInTempDir(prompt, effort, deps) {
299
308
  const runDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mandrel-code-review-'));
300
309
  try {
301
310
  const auth = readAuthStatus(await runAuthStatus(runDir, deps));
302
311
  if (!auth.ok) return auth;
303
312
  deps.logger?.info?.(
304
- `[code-review] Invoking claude --print --effort low (${auth.detail})...`,
313
+ `[code-review] Invoking claude --print --effort ${effort} (${auth.detail})...`,
305
314
  );
306
- return resolveVerdict(await runReviewer(runDir, prompt, deps));
315
+ return resolveVerdict(await runReviewer(runDir, prompt, effort, deps));
307
316
  } finally {
308
317
  fs.rmSync(runDir, { recursive: true, force: true });
309
318
  }
310
319
  }
311
320
 
321
+ /** Waits out the failure's `delayMs`, when it names one. */
322
+ async function pauseFor({ delayMs }, deps) {
323
+ if (delayMs) await (deps.sleepFn ?? defaultSleep)(delayMs);
324
+ }
325
+
326
+ /** One invocation, re-run once when its failure is transient-shaped. */
327
+ async function reviewWithRetry(prompt, effort, deps) {
328
+ const first = await reviewInTempDir(prompt, effort, deps);
329
+ if (!first.retryable) return first;
330
+ deps.logger?.warn?.(
331
+ `[code-review] No verdict (${first.reason}); retrying once.`,
332
+ );
333
+ await pauseFor(first, deps);
334
+ return reviewInTempDir(prompt, effort, deps);
335
+ }
336
+
337
+ /** @param {ReviewInput} input */
338
+ const effortOf = (input) => input.effort ?? 'low';
339
+
312
340
  function assertCliPresent(probeFn) {
313
341
  if (probeFn()) return;
314
342
  throw new Error(
@@ -325,6 +353,7 @@ function assertCliPresent(probeFn) {
325
353
  * gitSpawnFn?: typeof gitSpawn,
326
354
  * spawnFn?: Function,
327
355
  * authSpawnFn?: Function,
356
+ * sleepFn?: (ms: number) => Promise<void>,
328
357
  * platform?: string,
329
358
  * logger?: { info?: Function, warn?: Function },
330
359
  * }} [deps] - `spawnFn` / `authSpawnFn` replace the async runner for the
@@ -339,18 +368,38 @@ export function createCodeReviewProviderForRegistry(deps = {}) {
339
368
  let degradations = [];
340
369
 
341
370
  /** Record a not-checked slice without replacing the findings. */
342
- function noteDegradation(reason) {
343
- degradations.push({ ...DEGRADATION_BASE, reason });
371
+ function noteDegradation(reason, kind) {
372
+ degradations.push({ ...DEGRADATION_BASE, kind, reason });
344
373
  deps.logger?.warn?.(`[code-review] Not checked: ${reason}.`);
345
374
  }
346
375
 
347
376
  function settle(verdict) {
348
377
  if (!('reason' in verdict)) return verdict.findings;
349
- degradations.push({ ...DEGRADATION_BASE, reason: verdict.reason });
378
+ degradations.push({
379
+ ...DEGRADATION_BASE,
380
+ kind: 'verdict-missing',
381
+ reason: verdict.reason,
382
+ });
350
383
  deps.logger?.warn?.(`[code-review] No verdict: ${verdict.reason}.`);
351
384
  return [];
352
385
  }
353
386
 
387
+ /**
388
+ * No chunk with a verdict is a review with none: one record naming every
389
+ * file. Otherwise each chunk without one is a partial gap.
390
+ */
391
+ function noteChunkGaps(failed, total) {
392
+ if (failed.length > 0 && failed.length === total) {
393
+ const reasons = [...new Set(failed.map((f) => f.reason))].join('; ');
394
+ const files = failed.map((f) => f.files).join(', ');
395
+ noteDegradation(`no verdict for ${files}: ${reasons}`, 'verdict-missing');
396
+ return;
397
+ }
398
+ for (const { files, reason } of failed) {
399
+ noteDegradation(`no verdict for ${files}: ${reason}`, 'coverage-partial');
400
+ }
401
+ }
402
+
354
403
  /**
355
404
  * The `<callers>` block, guidance, the slice note and the manifest ride every chunk,
356
405
  * so they come out of the budget; chunk numbers are one digit, so the
@@ -376,15 +425,18 @@ export function createCodeReviewProviderForRegistry(deps = {}) {
376
425
  noteDegradation(
377
426
  `review bounded at ${MAX_REVIEW_CHUNKS} chunks: ${overflow.length} ` +
378
427
  `file(s) not reviewed: ${overflow.join(', ')}`,
428
+ 'coverage-partial',
379
429
  );
380
430
  }
381
431
  const findings = [];
432
+ const failed = [];
382
433
  for (const [index, chunk] of chunks.entries()) {
383
434
  const files = chunk.sections.map((s) => s.file).join(', ');
384
435
  if (chunk.size > budget) {
385
436
  noteDegradation(
386
437
  `${files} truncated at ${budget} characters: ` +
387
438
  `${chunk.size - budget} characters not reviewed`,
439
+ 'coverage-partial',
388
440
  );
389
441
  }
390
442
  const prompt = buildPrompt(
@@ -394,13 +446,11 @@ export function createCodeReviewProviderForRegistry(deps = {}) {
394
446
  budget,
395
447
  sliceNote(index + 1, chunks.length) + manifest,
396
448
  );
397
- const verdict = await reviewInTempDir(prompt, runDeps);
398
- if ('reason' in verdict) {
399
- noteDegradation(`no verdict for ${files}: ${verdict.reason}`);
400
- } else {
401
- findings.push(...verdict.findings);
402
- }
449
+ const verdict = await reviewWithRetry(prompt, effortOf(input), runDeps);
450
+ if ('reason' in verdict) failed.push({ files, reason: verdict.reason });
451
+ else findings.push(...verdict.findings);
403
452
  }
453
+ noteChunkGaps(failed, chunks.length);
404
454
  return dedupeFindings(findings);
405
455
  }
406
456
 
@@ -417,13 +467,14 @@ export function createCodeReviewProviderForRegistry(deps = {}) {
417
467
  cwd: PROJECT_ROOT,
418
468
  gitSpawnFn,
419
469
  });
420
- if (callers.reason) noteDegradation(callers.reason);
470
+ if (callers.reason) noteDegradation(callers.reason, 'context-reduced');
421
471
  if (diff.text.length > MAX_DIFF_CHARS) {
422
472
  return reviewInChunks(input, diff.text, callers.text);
423
473
  }
424
474
  return settle(
425
- await reviewInTempDir(
475
+ await reviewWithRetry(
426
476
  buildPrompt(input, diff.text, callers.text),
477
+ effortOf(input),
427
478
  runDeps,
428
479
  ),
429
480
  );
@@ -256,6 +256,18 @@ function boundDegradation(skipped) {
256
256
  ];
257
257
  }
258
258
 
259
+ function voterContext(args) {
260
+ return {
261
+ baseRef: args.baseRef,
262
+ headRef: args.headRef,
263
+ cwd: args.cwd ?? PROJECT_ROOT,
264
+ spawnFn: args.spawnFn,
265
+ platform: args.platform ?? process.platform,
266
+ gitSpawnFn: args.gitSpawnFn ?? gitSpawn,
267
+ cliPresent: (args.probeFn ?? probeClaudeCli)(),
268
+ };
269
+ }
270
+
259
271
  /**
260
272
  * Verify every `critical` finding; findings below critical are never sent.
261
273
  * One verifier at `light`/`standard`, three at `deep`.
@@ -278,15 +290,7 @@ async function verifyCriticals(args) {
278
290
  }
279
291
  const checked = criticals.slice(0, MAX_VERIFIED_CRITICALS);
280
292
  const voters = args.depth === 'deep' ? 3 : 1;
281
- const ctx = {
282
- baseRef: args.baseRef,
283
- headRef: args.headRef,
284
- cwd: args.cwd ?? PROJECT_ROOT,
285
- spawnFn: args.spawnFn,
286
- platform: args.platform ?? process.platform,
287
- gitSpawnFn: args.gitSpawnFn ?? gitSpawn,
288
- cliPresent: (args.probeFn ?? probeClaudeCli)(),
289
- };
293
+ const ctx = voterContext(args);
290
294
  args.logger?.info?.(
291
295
  `[code-review] Verifying ${checked.length} critical(s) with ${voters} verifier(s) each...`,
292
296
  );
@@ -313,6 +317,56 @@ async function verifyCriticals(args) {
313
317
  };
314
318
  }
315
319
 
320
+ /** Whether the review gets a verifier: built-in, or injected seams. */
321
+ const hasVerifier = (opts) => Boolean(opts.verifier || !opts.reviewProvider);
322
+
323
+ /**
324
+ * Re-check earlier CRITICALs at the current head: one call each whatever the
325
+ * depth, the same pool and cap as `verifyCriticals`. Each result is the voter
326
+ * outcome; past the cap or with no verifier the outcome is `skipped`.
327
+ *
328
+ * @param {{ findings: Finding[], reviewInput: { baseRef: string,
329
+ * headRef: string }, opts: object }} args
330
+ * @returns {Promise<{ results: { outcome: string, reason: string }[],
331
+ * degradations: GateDegradation[], verification: VerificationSummary }>}
332
+ */
333
+ export async function recheckCriticals({ findings, reviewInput, opts }) {
334
+ const verification = emptyVerification();
335
+ if (findings.length === 0 || !hasVerifier(opts)) {
336
+ const reason = 'no verifier for this review';
337
+ return {
338
+ results: findings.map(() => ({ outcome: 'skipped', reason })),
339
+ degradations: [],
340
+ verification,
341
+ };
342
+ }
343
+ const checked = findings.slice(0, MAX_VERIFIED_CRITICALS);
344
+ const ctx = voterContext({
345
+ baseRef: reviewInput.baseRef,
346
+ headRef: reviewInput.headRef,
347
+ cwd: opts.cwd,
348
+ gitSpawnFn: opts.gitSpawnFn,
349
+ ...opts.verifier,
350
+ });
351
+ const votes = await runPooled(
352
+ checked.map((f) => () => runVoter(f, ctx)),
353
+ MAX_IN_FLIGHT,
354
+ );
355
+ for (const v of votes) verification[v.outcome] += 1;
356
+ verification.attempted = votes.length;
357
+ return {
358
+ results: findings.map(
359
+ (_, i) =>
360
+ votes[i] ?? { outcome: 'skipped', reason: 'verification cap reached' },
361
+ ),
362
+ degradations: [
363
+ ...checked.flatMap((f, i) => voteDegradations(f, [votes[i]])),
364
+ ...boundDegradation(findings.slice(MAX_VERIFIED_CRITICALS)),
365
+ ],
366
+ verification,
367
+ };
368
+ }
369
+
316
370
  /**
317
371
  * The pipeline step: ground every finding, then verify the CRITICALs left.
318
372
  * An injected review provider gets no verifier unless `opts.verifier` (the
@@ -337,8 +391,7 @@ export async function vetFindings({
337
391
  degradations,
338
392
  gitSpawnFn: opts.gitSpawnFn,
339
393
  });
340
- const verify =
341
- opts.verifier || !opts.reviewProvider ? verifyCriticals : skipVerification;
394
+ const verify = hasVerifier(opts) ? verifyCriticals : skipVerification;
342
395
  const verified = await verify({
343
396
  findings: grounded.findings,
344
397
  baseRef,
@@ -1,17 +1,29 @@
1
1
  /**
2
2
  * review-providers/degraded-gates.js — the "a gate did not run" channel,
3
- * travelling beside `Finding[]` so "never ran" cannot read as clean. It
4
- * reports rather than blocks: close already hard-gates `npm run lint`.
3
+ * travelling beside `Finding[]` so "never ran" cannot read as clean. Its
4
+ * `kind` blocks only under `delivery.codeReview.onNoVerdict: "block"`.
5
5
  */
6
6
 
7
7
  /**
8
+ * @typedef {'verdict-missing'|'coverage-partial'|'context-reduced'} DegradationKind
8
9
  * @typedef {object} GateDegradation
9
10
  * @property {string} tool
10
11
  * @property {string} gate
11
12
  * @property {string} surface Sub-surface that could not run.
13
+ * @property {DegradationKind} kind
12
14
  * @property {string} reason
13
15
  */
14
16
 
17
+ const DEGRADATION_KINDS = Object.freeze([
18
+ 'verdict-missing',
19
+ 'coverage-partial',
20
+ 'context-reduced',
21
+ ]);
22
+
23
+ /** An absent or unknown kind is the one kind that can never block. */
24
+ const kindOf = (kind) =>
25
+ DEGRADATION_KINDS.includes(kind) ? kind : 'context-reduced';
26
+
15
27
  /**
16
28
  * Keep only well-formed records, so a misbehaving provider cannot corrupt output.
17
29
  *
@@ -30,12 +42,99 @@ export function normalizeDegradations(input) {
30
42
  tool: typeof d.tool === 'string' ? d.tool : 'unknown',
31
43
  gate: typeof d.gate === 'string' ? d.gate : 'unknown',
32
44
  surface,
45
+ kind: kindOf(d.kind),
33
46
  reason,
34
47
  });
35
48
  }
36
49
  return out;
37
50
  }
38
51
 
52
+ /**
53
+ * Close's `gates.codeReview`: overridden > unreviewed > partial > degraded > passed.
54
+ *
55
+ * @param {unknown} degradations
56
+ * @param {{ overridden?: boolean }} [opts]
57
+ * @returns {'overridden'|'unreviewed'|'partial'|'degraded'|'passed'}
58
+ */
59
+ export function reviewGateValue(degradations, { overridden = false } = {}) {
60
+ if (overridden) return 'overridden';
61
+ const kinds = new Set(normalizeDegradations(degradations).map((d) => d.kind));
62
+ if (kinds.has('verdict-missing')) return 'unreviewed';
63
+ if (kinds.has('coverage-partial')) return 'partial';
64
+ return kinds.size > 0 ? 'degraded' : 'passed';
65
+ }
66
+
67
+ const REDUCED_CONTEXT =
68
+ '⚠️ Review ran DEGRADED — reviewed with reduced context, so the surface(s) ' +
69
+ 'above were not fully checked. The close is not blocked by reduced ' +
70
+ 'context, but nothing here vouches for those surfaces.';
71
+
72
+ const NOT_VOUCHED =
73
+ 'unless `delivery.codeReview.onNoVerdict` is `block`; nothing here vouches ' +
74
+ 'for those surfaces.';
75
+
76
+ /**
77
+ * @param {unknown} degradations
78
+ * @returns {string[]}
79
+ */
80
+ export function reviewGapLines(degradations) {
81
+ const gate = reviewGateValue(degradations);
82
+ if (gate === 'unreviewed') {
83
+ return [
84
+ `⚠️ Review UNREVIEWED — no verdict for the surface(s) above. The close is not blocked ${NOT_VOUCHED}`,
85
+ ];
86
+ }
87
+ if (gate === 'partial') {
88
+ return [
89
+ `⚠️ Review ran PARTIAL — part of the diff above was not reviewed. The close is not blocked ${NOT_VOUCHED}`,
90
+ ];
91
+ }
92
+ return gate === 'degraded' ? [REDUCED_CONTEXT] : [];
93
+ }
94
+
95
+ /** The fix each no-verdict cause needs; matched on the record, never on PII. */
96
+ const NO_VERDICT_FIXES = [
97
+ [
98
+ /auth status returned unparseable output/,
99
+ 'the auth preflight output was transient; re-run `story-review-compute.js`',
100
+ ],
101
+ [/not logged in|auth status/, 'run `claude auth login`'],
102
+ [/out of quota|usage limit/, 'wait for the usage-limit reset'],
103
+ [/skipped/, 'install the `claude` or `codex` CLI'],
104
+ [/remote-base-ref-unresolved/, 'fetch the base ref'],
105
+ ];
106
+
107
+ function fixFor(reason) {
108
+ const hit = NO_VERDICT_FIXES.find(([re]) => re.test(reason));
109
+ return hit ? hit[1] : 'fix the cause';
110
+ }
111
+
112
+ /**
113
+ * Under `onNoVerdict: "block"`, a verdict-missing or coverage-partial record
114
+ * halts the review; context-reduced never does.
115
+ *
116
+ * @param {{ degradations: unknown, onNoVerdict?: string, storyId: number|string }} args
117
+ * @returns {{ halted: true, haltCause: 'no-verdict', blockerReason: string }|null}
118
+ */
119
+ export function resolveNoVerdictHalt({ degradations, onNoVerdict, storyId }) {
120
+ if (onNoVerdict !== 'block') return null;
121
+ const blocking = normalizeDegradations(degradations).filter(
122
+ (d) => d.kind !== 'context-reduced',
123
+ );
124
+ if (blocking.length === 0) return null;
125
+ const causes = blocking.map(
126
+ (d) => `${d.surface} [${d.kind}]: ${d.reason} — ${fixFor(d.reason)}`,
127
+ );
128
+ return {
129
+ halted: true,
130
+ haltCause: 'no-verdict',
131
+ blockerReason:
132
+ `code-review produced no verdict for part or all of the diff ` +
133
+ `(delivery.codeReview.onNoVerdict: block): ${causes.join('; ')}. ` +
134
+ `Then re-run \`node .agents/scripts/story-review-compute.js --story ${storyId}\`.`,
135
+ };
136
+ }
137
+
39
138
  /**
40
139
  * Feature-detected; a throw degrades to empty. MUST be called after
41
140
  * `runReview`, which is when a provider records its degradations.
@@ -152,7 +251,7 @@ export function renderDegradedGatesSection(degradations) {
152
251
  ];
153
252
  for (const d of rows) {
154
253
  lines.push(
155
- `- \`${d.gate}\` → \`${d.surface}\` could not execute — ${d.reason} (emitter: \`${d.tool}\`).`,
254
+ `- \`${d.gate}\` → \`${d.surface}\` (${d.kind}) could not execute — ${d.reason} (emitter: \`${d.tool}\`).`,
156
255
  );
157
256
  }
158
257
  lines.push('');
@@ -34,12 +34,16 @@ function normalizeTitle(title) {
34
34
 
35
35
  /**
36
36
  * First 12 hex chars of sha256 over class (else category), first instance
37
- * file and, for a class-less finding, its normalized title.
37
+ * file and, for a class-less finding, its normalized title. A finding carried
38
+ * forward from the review history keeps the id it was recorded under.
38
39
  *
39
40
  * @param {Finding} finding
40
41
  * @returns {string}
41
42
  */
42
- export function findingId(finding) {
43
+ export const findingId = (finding) => finding?.carriedId ?? contentId(finding);
44
+
45
+ /** @param {Finding} finding */
46
+ function contentId(finding) {
43
47
  const cls = finding?.class;
44
48
  const parts = [
45
49
  cls || finding?.category || 'bug',