mandrel 2.19.0 → 2.20.0
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.
- package/.agents/docs/agentrc-reference.json +0 -21
- package/.agents/docs/configuration.md +11 -14
- package/.agents/docs/execution-reference.md +8 -5
- package/.agents/schemas/agentrc.schema.json +0 -31
- package/.agents/scripts/check-test-temp-hygiene.js +153 -14
- package/.agents/scripts/lib/baselines/kinds/maintainability.js +3 -4
- package/.agents/scripts/lib/bdd-scenario-scanner.js +3 -2
- package/.agents/scripts/lib/config/explain.js +0 -8
- package/.agents/scripts/lib/config/temp-paths.js +12 -1
- package/.agents/scripts/lib/config-settings-schema.js +8 -24
- package/.agents/scripts/lib/orchestration/file-assumptions.js +4 -2
- package/.agents/scripts/lib/orchestration/plan-context.js +13 -14
- package/.agents/scripts/lib/orchestration/planning/authoring-context.js +12 -66
- package/.agents/scripts/lib/test-env.js +15 -3
- package/.agents/scripts/lib/test-temp.js +311 -0
- package/.agents/workflows/helpers/plan-reference.md +19 -4
- package/.agents/workflows/plan.md +7 -6
- package/docs/CHANGELOG.md +8 -0
- package/lib/migrations/index.js +2 -0
- package/lib/migrations/steps/2.20.0-retire-codebase-snapshot.js +113 -0
- package/package.json +1 -1
- package/.agents/scripts/lib/codebase-snapshot.js +0 -513
- package/.agents/scripts/lib/orchestration/planning/spec-authoring-grounding.js +0 -147
- package/.agents/scripts/lib/orchestration/spec-freshness.js +0 -129
|
@@ -67,27 +67,6 @@
|
|
|
67
67
|
}
|
|
68
68
|
},
|
|
69
69
|
"planning": {
|
|
70
|
-
"codebaseSnapshot": {
|
|
71
|
-
"tier": "skinny",
|
|
72
|
-
"include": [
|
|
73
|
-
".agents/scripts/**",
|
|
74
|
-
"src/**",
|
|
75
|
-
"lib/**",
|
|
76
|
-
"app/**",
|
|
77
|
-
"packages/**"
|
|
78
|
-
],
|
|
79
|
-
"exclude": [
|
|
80
|
-
"**/node_modules/**",
|
|
81
|
-
"**/dist/**",
|
|
82
|
-
"**/build/**",
|
|
83
|
-
"**/.next/**",
|
|
84
|
-
"**/.turbo/**",
|
|
85
|
-
"**/coverage/**",
|
|
86
|
-
"**/*.test.*",
|
|
87
|
-
"**/*.spec.*"
|
|
88
|
-
],
|
|
89
|
-
"recentCommitWindow": 30
|
|
90
|
-
},
|
|
91
70
|
"riskHeuristics": [
|
|
92
71
|
"Destructive or irreversible data mutations (dropping tables, deleting rows without soft-delete or backup, truncating production state).",
|
|
93
72
|
"Modifications to shared security or auth infrastructure (IAM policies, auth middleware, session or token handling, secret rotation).",
|
|
@@ -26,7 +26,7 @@ with the runtime validators.
|
|
|
26
26
|
"$schema": "./.agents/schemas/agentrc.schema.json",
|
|
27
27
|
"project": { /* paths, commands, baseBranch, docsContextFiles */ },
|
|
28
28
|
"github": { /* owner, repo, branchProtection, mergeMethods, notifications */ },
|
|
29
|
-
"planning": { /* riskHeuristics, conflict gates,
|
|
29
|
+
"planning": { /* riskHeuristics, conflict gates, complexityGate, navigation */ },
|
|
30
30
|
"delivery": { /* execution, quality, worktreeIsolation, deliverRunner, ... */ }
|
|
31
31
|
}
|
|
32
32
|
```
|
|
@@ -38,7 +38,7 @@ top-level keys are validation errors.
|
|
|
38
38
|
| ------------- | -------- | ---------------------------------------------------------------------------------- |
|
|
39
39
|
| `project` | **Yes** | Project-local paths, base branch, validation commands, and context-hydration files. |
|
|
40
40
|
| `github` | No | Ticketing provider config: owner/repo, branch protection, merge methods, notifications. |
|
|
41
|
-
| `planning` | No | `/plan` tuning: conflict advisories,
|
|
41
|
+
| `planning` | No | `/plan` tuning: conflict advisories, complexity routing, navigability gate. (Story sizing ceilings and the planner-context cap are code-absolute — not agentrc.) |
|
|
42
42
|
| `delivery` | No | `/deliver` tuning: quality gates, worktree isolation, runners, CI watch, code-review providers. |
|
|
43
43
|
| `$schema` | No | JSON Schema pointer for editor tooling. |
|
|
44
44
|
|
|
@@ -97,11 +97,6 @@ top-level keys are validation errors.
|
|
|
97
97
|
| Key | Required | Type | Default | Description |
|
|
98
98
|
| --- | --- | --- | --- | --- |
|
|
99
99
|
| `riskHeuristics` | No | `string[]` or `{ append?, prepend? }` | — | — |
|
|
100
|
-
| `codebaseSnapshot` | No | `object` | — | Nested configuration block. |
|
|
101
|
-
| `codebaseSnapshot.tier` | No | `"skinny"` \| `"medium"` | — | — |
|
|
102
|
-
| `codebaseSnapshot.include` | No | `array<string>` | — | — |
|
|
103
|
-
| `codebaseSnapshot.exclude` | No | `array<string>` | — | — |
|
|
104
|
-
| `codebaseSnapshot.recentCommitWindow` | No | `integer` | — | — |
|
|
105
100
|
| `complexityGate` | No | `object` | — | Shape-derived ceremony-lite complexity routing. A lite claim is validated against the authored Story shape at persist and re-derived from the Story body at dispatch; conservative (full on any doubt). Never relaxes the Story-ticket / PR-to-main / repo-gates / security-baseline non-negotiables. |
|
|
106
101
|
| `complexityGate.enabled` | No | `boolean` | — | Master switch. When false, lite routing is disabled everywhere: persist refuses lite claims and dispatch always takes the sub-agent path. Default true. |
|
|
107
102
|
| `complexityGate.maxArtifacts` | No | `integer` | — | Enumerated-artifact threshold reported by the plan-context complexity signals. An input signal for the planner verdict — carries no routing authority. Default 1. |
|
|
@@ -352,13 +347,15 @@ Defaults are **advisory, not Story-width ceilings**. Session-mass ceilings live
|
|
|
352
347
|
as absolute authored-token constants on `DEFAULT_MODEL_CAPACITY` in
|
|
353
348
|
`ticket-validator-sizing.js` (soft 30k / hard 75k); there is no `maxTokenBudget`
|
|
354
349
|
envelope, and cohesion / split policy / conflict advisories are the primary
|
|
355
|
-
sizing signal.
|
|
356
|
-
|
|
357
|
-
|
|
358
|
-
|
|
359
|
-
|
|
360
|
-
`
|
|
361
|
-
|
|
350
|
+
sizing signal.
|
|
351
|
+
|
|
352
|
+
There is no snapshot knob to tune: Story #4811 retired the structural-snapshot
|
|
353
|
+
block outright. Spec authoring is grounded by the author's own targeted repo
|
|
354
|
+
retrieval plus the Phase 8 `validateStoryFileAssumptions` gate, which probes
|
|
355
|
+
every authored `{path, assumption}` against the real tree as a hard error — a
|
|
356
|
+
pre-computed inventory added a second, staler answer to the same question.
|
|
357
|
+
A config still carrying the retired key is a hard validation failure; the
|
|
358
|
+
2.20.0 retirement migration strips it on upgrade.
|
|
362
359
|
|
|
363
360
|
- **`complexityGate`.** Shape-derived ceremony-lite routing (Story #4722,
|
|
364
361
|
superseding the word-count gate of Stories #4683/#4707). The full ceremony
|
|
@@ -96,12 +96,15 @@ over-ceiling envelope or an over-budget Story count.
|
|
|
96
96
|
- **`PLAN_CONTEXT_ENVELOPE_BYTE_CEILING`** (`lib/orchestration/plan-context.js`):
|
|
97
97
|
256 KB (≈64K tokens at the ≈4-chars/token estimate) on the serialized
|
|
98
98
|
envelope `buildPlanContext` assembles, checked at the single choke point
|
|
99
|
-
every mode returns through.
|
|
100
|
-
|
|
101
|
-
|
|
99
|
+
every mode returns through. A measured seed-mode envelope on this repo is
|
|
100
|
+
~120 KB — `docsContext` (~63 KB) and `systemPrompts` (~54 KB) are the whole
|
|
101
|
+
of it, every other field under 1 KB — so the ceiling carries roughly 2×
|
|
102
|
+
headroom. Story #4811 retired the codebase snapshot that used to sit
|
|
103
|
+
alongside them, leaving the operator-supplied seed the only unbounded
|
|
104
|
+
contributor.
|
|
102
105
|
- **On refusal**, the error names the envelope's largest fields. Trim the seed,
|
|
103
|
-
plan fewer `--tickets` source issues in one run
|
|
104
|
-
|
|
106
|
+
or plan fewer `--tickets` source issues in one run. The seed is carried
|
|
107
|
+
**verbatim** by design — it
|
|
105
108
|
is the operator's request, and summarizing it silently would degrade planning
|
|
106
109
|
quality precisely when the input is richest — so there is no elision path to
|
|
107
110
|
fall back on. Raising the ceiling needs a measured justification.
|
|
@@ -324,9 +324,6 @@
|
|
|
324
324
|
"riskHeuristics": {
|
|
325
325
|
"$ref": "#/$defs/listOrExtenderOfStrings"
|
|
326
326
|
},
|
|
327
|
-
"codebaseSnapshot": {
|
|
328
|
-
"$ref": "#/$defs/codebaseSnapshot"
|
|
329
|
-
},
|
|
330
327
|
"complexityGate": {
|
|
331
328
|
"type": "object",
|
|
332
329
|
"description": "Shape-derived ceremony-lite complexity routing. A lite claim is validated against the authored Story shape at persist and re-derived from the Story body at dispatch; conservative (full on any doubt). Never relaxes the Story-ticket / PR-to-main / repo-gates / security-baseline non-negotiables.",
|
|
@@ -392,34 +389,6 @@
|
|
|
392
389
|
},
|
|
393
390
|
"additionalProperties": false
|
|
394
391
|
},
|
|
395
|
-
"codebaseSnapshot": {
|
|
396
|
-
"type": "object",
|
|
397
|
-
"properties": {
|
|
398
|
-
"tier": {
|
|
399
|
-
"type": "string",
|
|
400
|
-
"enum": ["skinny", "medium"]
|
|
401
|
-
},
|
|
402
|
-
"include": {
|
|
403
|
-
"type": "array",
|
|
404
|
-
"items": {
|
|
405
|
-
"type": "string",
|
|
406
|
-
"minLength": 1
|
|
407
|
-
}
|
|
408
|
-
},
|
|
409
|
-
"exclude": {
|
|
410
|
-
"type": "array",
|
|
411
|
-
"items": {
|
|
412
|
-
"type": "string",
|
|
413
|
-
"minLength": 1
|
|
414
|
-
}
|
|
415
|
-
},
|
|
416
|
-
"recentCommitWindow": {
|
|
417
|
-
"type": "integer",
|
|
418
|
-
"minimum": 1
|
|
419
|
-
}
|
|
420
|
-
},
|
|
421
|
-
"additionalProperties": false
|
|
422
|
-
},
|
|
423
392
|
"execution": {
|
|
424
393
|
"type": "object",
|
|
425
394
|
"properties": {
|
|
@@ -15,16 +15,41 @@
|
|
|
15
15
|
* per-process scratch dir. This script is the regression guard that keeps
|
|
16
16
|
* the fix honest, plus a local cleanup mode for the accumulated noise:
|
|
17
17
|
*
|
|
18
|
+
* The guard covers two distinct temp roots, and conflating them is how the
|
|
19
|
+
* second one went unmeasured for so long:
|
|
20
|
+
*
|
|
21
|
+
* 1. The repo's own `temp/` telemetry tree — the original dimension above.
|
|
22
|
+
* 2. The **OS temp root** (Story #4808). The redirect in (1) sends stray
|
|
23
|
+
* writes into `os.tmpdir()` scratch dirs, and nothing ever reaped them:
|
|
24
|
+
* the remedy for (1) became the largest single leaker into (2). Since
|
|
25
|
+
* the damaging axis there is entry *count*, the suite now nests every
|
|
26
|
+
* managed dir inside one per-process `mandrel-suite-*` root
|
|
27
|
+
* (`lib/test-temp.js`) and reaps it, and this guard asserts that no
|
|
28
|
+
* such root survives a run.
|
|
29
|
+
*
|
|
18
30
|
* --snapshot Record a fingerprint (size + sha256) of every stream
|
|
19
|
-
* file under `temp
|
|
20
|
-
*
|
|
31
|
+
* file under `temp/`, plus the `mandrel-suite-*` roots
|
|
32
|
+
* already present in the OS temp root, to the snapshot
|
|
33
|
+
* baseline. Run this before the suite.
|
|
21
34
|
* --assert Re-scan and fail if any stream file was added or grew
|
|
22
|
-
* relative to the snapshot
|
|
23
|
-
*
|
|
35
|
+
* relative to the snapshot, or if a suite root appeared
|
|
36
|
+
* and survived. Run this after the suite. A missing
|
|
37
|
+
* snapshot is a hard failure ("snapshot missing
|
|
24
38
|
* — guard cannot attest"), never a silent re-baseline:
|
|
25
39
|
* the baseline lives *outside* the protected `temp/`
|
|
26
40
|
* tree (Story #4711), so a test wiping `temp/` can no
|
|
27
41
|
* longer destroy the baseline and fail the guard open.
|
|
42
|
+
* Recording pre-existing suite roots (rather than
|
|
43
|
+
* asserting an empty set) is what keeps a concurrent
|
|
44
|
+
* suite in another checkout from failing this one.
|
|
45
|
+
* --lint-globs <g> Comma-separated repo-relative globs to scan for test
|
|
46
|
+
* files that call `mkdtemp` against `os.tmpdir()`
|
|
47
|
+
* directly instead of going through `makeTempDir`.
|
|
48
|
+
* **Off unless passed**: this script ships in the
|
|
49
|
+
* materialized `.agents/` payload and a consumer's
|
|
50
|
+
* tests are none of this rule's business. A line (or
|
|
51
|
+
* the line above it) carrying `test-temp-allow` opts
|
|
52
|
+
* out.
|
|
28
53
|
* --baseline <path> Explicit snapshot-baseline path (CI sets this to a
|
|
29
54
|
* runner-temp path). Defaults to an OS scratch location
|
|
30
55
|
* keyed by the resolved repo root. Refused when it
|
|
@@ -53,6 +78,12 @@ import os from 'node:os';
|
|
|
53
78
|
import path from 'node:path';
|
|
54
79
|
import { fileURLToPath } from 'node:url';
|
|
55
80
|
import { runAsCli } from './lib/cli-utils.js';
|
|
81
|
+
import {
|
|
82
|
+
findRawTmpdirMkdtemp,
|
|
83
|
+
listSuiteTempRoots,
|
|
84
|
+
SUITE_ROOTS_KEY,
|
|
85
|
+
survivingSuiteTempRoots,
|
|
86
|
+
} from './lib/test-temp.js';
|
|
56
87
|
|
|
57
88
|
const __dirname = path.dirname(fileURLToPath(import.meta.url));
|
|
58
89
|
const REPO_ROOT = path.resolve(__dirname, '..', '..');
|
|
@@ -193,17 +224,27 @@ export function buildManifest(tempDir) {
|
|
|
193
224
|
* `defaultBaselinePath` — never inside `temp/`).
|
|
194
225
|
* @param {string} repoRoot
|
|
195
226
|
* @param {string} [baselinePath]
|
|
196
|
-
* @
|
|
227
|
+
* @param {{ tmpDir?: string }} [deps] Injectable OS temp root for tests.
|
|
228
|
+
* @returns {{ snapshotPath: string, count: number, suiteRoots: number }}
|
|
197
229
|
*/
|
|
198
|
-
export function writeSnapshot(repoRoot, baselinePath) {
|
|
230
|
+
export function writeSnapshot(repoRoot, baselinePath, { tmpDir } = {}) {
|
|
199
231
|
const snapshotPath = checkedBaselinePath(
|
|
200
232
|
repoRoot,
|
|
201
233
|
baselinePath ?? defaultBaselinePath(repoRoot),
|
|
202
234
|
);
|
|
203
235
|
const manifest = buildManifest(tempDirFor(repoRoot));
|
|
236
|
+
const count = Object.keys(manifest).length;
|
|
237
|
+
// Reserved key: stream entries are always `*.ndjson` relative paths, so
|
|
238
|
+
// this cannot shadow one, and `diffAgainstSnapshot` only ever looks up
|
|
239
|
+
// keys derived from the tree it just walked.
|
|
240
|
+
manifest[SUITE_ROOTS_KEY] = listSuiteTempRoots(tmpDir ?? os.tmpdir());
|
|
204
241
|
mkdirSync(path.dirname(snapshotPath), { recursive: true });
|
|
205
242
|
writeFileSync(snapshotPath, `${JSON.stringify(manifest, null, 2)}\n`, 'utf8');
|
|
206
|
-
return {
|
|
243
|
+
return {
|
|
244
|
+
snapshotPath,
|
|
245
|
+
count,
|
|
246
|
+
suiteRoots: manifest[SUITE_ROOTS_KEY].length,
|
|
247
|
+
};
|
|
207
248
|
}
|
|
208
249
|
|
|
209
250
|
/**
|
|
@@ -331,7 +372,7 @@ export function cleanFixtureDirs({
|
|
|
331
372
|
/**
|
|
332
373
|
* Parse the CLI argv into a normalised options object.
|
|
333
374
|
* @param {string[]} argv
|
|
334
|
-
* @returns {{ mode: 'snapshot'|'assert'|'clean', apply: boolean, ids: number[]|null, repoRoot: string, baseline: string|null }}
|
|
375
|
+
* @returns {{ mode: 'snapshot'|'assert'|'clean', apply: boolean, ids: number[]|null, repoRoot: string, baseline: string|null, lintGlobs: string[] }}
|
|
335
376
|
*/
|
|
336
377
|
export function parseArgv(argv) {
|
|
337
378
|
let mode = 'assert';
|
|
@@ -339,13 +380,20 @@ export function parseArgv(argv) {
|
|
|
339
380
|
let ids = null;
|
|
340
381
|
let repoRoot = REPO_ROOT;
|
|
341
382
|
let baseline = null;
|
|
383
|
+
let lintGlobs = [];
|
|
342
384
|
for (let i = 0; i < argv.length; i += 1) {
|
|
343
385
|
const arg = argv[i];
|
|
344
386
|
if (arg === '--snapshot') mode = 'snapshot';
|
|
345
387
|
else if (arg === '--assert') mode = 'assert';
|
|
346
388
|
else if (arg === '--clean') mode = 'clean';
|
|
347
389
|
else if (arg === '--yes') apply = true;
|
|
348
|
-
else if (arg === '--
|
|
390
|
+
else if (arg === '--lint-globs') {
|
|
391
|
+
i += 1;
|
|
392
|
+
lintGlobs = String(argv[i] ?? '')
|
|
393
|
+
.split(',')
|
|
394
|
+
.map((s) => s.trim())
|
|
395
|
+
.filter(Boolean);
|
|
396
|
+
} else if (arg === '--ids') {
|
|
349
397
|
i += 1;
|
|
350
398
|
ids = String(argv[i] ?? '')
|
|
351
399
|
.split(',')
|
|
@@ -359,7 +407,7 @@ export function parseArgv(argv) {
|
|
|
359
407
|
baseline = path.resolve(String(argv[i] ?? '.'));
|
|
360
408
|
}
|
|
361
409
|
}
|
|
362
|
-
return { mode, apply, ids, repoRoot, baseline };
|
|
410
|
+
return { mode, apply, ids, repoRoot, baseline, lintGlobs };
|
|
363
411
|
}
|
|
364
412
|
|
|
365
413
|
/**
|
|
@@ -369,14 +417,23 @@ export function parseArgv(argv) {
|
|
|
369
417
|
*
|
|
370
418
|
* @param {ReturnType<typeof parseArgv>} opts
|
|
371
419
|
* @param {(line: string) => void} [log]
|
|
420
|
+
* @param {{ tmpDir?: string }} [deps] Injectable OS temp root for tests.
|
|
372
421
|
* @returns {number}
|
|
373
422
|
*/
|
|
374
|
-
export function runHygiene(
|
|
375
|
-
|
|
423
|
+
export function runHygiene(
|
|
424
|
+
opts,
|
|
425
|
+
log = (l) => process.stdout.write(`${l}\n`),
|
|
426
|
+
{ tmpDir = os.tmpdir() } = {},
|
|
427
|
+
) {
|
|
428
|
+
const { mode, apply, ids, repoRoot, baseline = null, lintGlobs = [] } = opts;
|
|
376
429
|
if (mode === 'snapshot') {
|
|
377
|
-
const { snapshotPath, count } = writeSnapshot(
|
|
430
|
+
const { snapshotPath, count, suiteRoots } = writeSnapshot(
|
|
431
|
+
repoRoot,
|
|
432
|
+
baseline,
|
|
433
|
+
{ tmpDir },
|
|
434
|
+
);
|
|
378
435
|
log(
|
|
379
|
-
`[test-temp-hygiene] snapshot recorded (${count} stream file(s)) → ${snapshotPath}`,
|
|
436
|
+
`[test-temp-hygiene] snapshot recorded (${count} stream file(s), ${suiteRoots} pre-existing suite root(s)) → ${snapshotPath}`,
|
|
380
437
|
);
|
|
381
438
|
return 0;
|
|
382
439
|
}
|
|
@@ -409,6 +466,26 @@ export function runHygiene(opts, log = (l) => process.stdout.write(`${l}\n`)) {
|
|
|
409
466
|
);
|
|
410
467
|
return 1;
|
|
411
468
|
}
|
|
469
|
+
// Every dimension runs and reports; a failure in one must not hide a
|
|
470
|
+
// failure in another, so the exit code is the max rather than an
|
|
471
|
+
// early return.
|
|
472
|
+
const codes = [
|
|
473
|
+
assertStreamTree(repoRoot, snapshot, log),
|
|
474
|
+
assertNoSurvivingSuiteRoots(snapshot, log, tmpDir),
|
|
475
|
+
assertNoRawTmpdirMkdtemp(repoRoot, lintGlobs, log),
|
|
476
|
+
];
|
|
477
|
+
return Math.max(...codes);
|
|
478
|
+
}
|
|
479
|
+
|
|
480
|
+
/**
|
|
481
|
+
* Dimension 1 — the repo's own `temp/` telemetry tree (Story #4696).
|
|
482
|
+
*
|
|
483
|
+
* @param {string} repoRoot
|
|
484
|
+
* @param {Record<string, unknown>} snapshot
|
|
485
|
+
* @param {(line: string) => void} log
|
|
486
|
+
* @returns {number} exit code
|
|
487
|
+
*/
|
|
488
|
+
function assertStreamTree(repoRoot, snapshot, log) {
|
|
412
489
|
const { added, changed } = diffAgainstSnapshot(
|
|
413
490
|
tempDirFor(repoRoot),
|
|
414
491
|
snapshot,
|
|
@@ -428,6 +505,68 @@ export function runHygiene(opts, log = (l) => process.stdout.write(`${l}\n`)) {
|
|
|
428
505
|
return 1;
|
|
429
506
|
}
|
|
430
507
|
|
|
508
|
+
/**
|
|
509
|
+
* Dimension 2 — the OS temp root (Story #4808). Fails when a suite root
|
|
510
|
+
* appeared since the snapshot and is still on disk, which means the run
|
|
511
|
+
* created it and never reaped it.
|
|
512
|
+
*
|
|
513
|
+
* @param {Record<string, unknown>} snapshot
|
|
514
|
+
* @param {(line: string) => void} log
|
|
515
|
+
* @param {string} tmpDir
|
|
516
|
+
* @returns {number} exit code
|
|
517
|
+
*/
|
|
518
|
+
function assertNoSurvivingSuiteRoots(snapshot, log, tmpDir) {
|
|
519
|
+
const before = Array.isArray(snapshot[SUITE_ROOTS_KEY])
|
|
520
|
+
? snapshot[SUITE_ROOTS_KEY]
|
|
521
|
+
: [];
|
|
522
|
+
const surviving = survivingSuiteTempRoots(tmpDir, before);
|
|
523
|
+
if (surviving.length === 0) {
|
|
524
|
+
log('[test-temp-hygiene] OK — no suite temp roots survived the run.');
|
|
525
|
+
return 0;
|
|
526
|
+
}
|
|
527
|
+
log(
|
|
528
|
+
`[test-temp-hygiene] FAIL — ${surviving.length} suite temp root(s) survived in ${tmpDir}:`,
|
|
529
|
+
);
|
|
530
|
+
for (const name of surviving) log(` + leaked ${name}`);
|
|
531
|
+
log(
|
|
532
|
+
'[test-temp-hygiene] a process minted a suite root and exited without reaping it. Do not delete these by hand — find the writer that bypassed makeTempDir().',
|
|
533
|
+
);
|
|
534
|
+
return 1;
|
|
535
|
+
}
|
|
536
|
+
|
|
537
|
+
/**
|
|
538
|
+
* Dimension 3 — the static backstop (Story #4808). Skipped, and reported
|
|
539
|
+
* as skipped, unless the caller passed `--lint-globs`.
|
|
540
|
+
*
|
|
541
|
+
* @param {string} repoRoot
|
|
542
|
+
* @param {string[]} globs
|
|
543
|
+
* @param {(line: string) => void} log
|
|
544
|
+
* @returns {number} exit code
|
|
545
|
+
*/
|
|
546
|
+
function assertNoRawTmpdirMkdtemp(repoRoot, globs, log) {
|
|
547
|
+
if (!globs || globs.length === 0) {
|
|
548
|
+
log(
|
|
549
|
+
'[test-temp-hygiene] SKIP — raw-tmpdir lint not requested (pass --lint-globs to enable).',
|
|
550
|
+
);
|
|
551
|
+
return 0;
|
|
552
|
+
}
|
|
553
|
+
const findings = findRawTmpdirMkdtemp(repoRoot, globs);
|
|
554
|
+
if (findings.length === 0) {
|
|
555
|
+
log(
|
|
556
|
+
'[test-temp-hygiene] OK — no test file mints OS temp dirs outside makeTempDir().',
|
|
557
|
+
);
|
|
558
|
+
return 0;
|
|
559
|
+
}
|
|
560
|
+
log(
|
|
561
|
+
`[test-temp-hygiene] FAIL — ${findings.length} raw os.tmpdir() mkdtemp call(s) in test files:`,
|
|
562
|
+
);
|
|
563
|
+
for (const f of findings) log(` ${f.file}:${f.line} ${f.text}`);
|
|
564
|
+
log(
|
|
565
|
+
"[test-temp-hygiene] use makeTempDir() from .agents/scripts/lib/test-temp.js so teardown is registered, or mark the line 'test-temp-allow: <reason>' when the real root is genuinely required.",
|
|
566
|
+
);
|
|
567
|
+
return 1;
|
|
568
|
+
}
|
|
569
|
+
|
|
431
570
|
runAsCli(
|
|
432
571
|
import.meta.url,
|
|
433
572
|
async () => {
|
|
@@ -60,11 +60,10 @@ export const MAINTAINABILITY_EXCLUSIONS = Object.freeze(
|
|
|
60
60
|
// the audit-to-stories parser reuses the same regex-property scan
|
|
61
61
|
// patterns as acceptance-spec-reconciler.
|
|
62
62
|
'.agents/scripts/lib/audit-to-stories/parse-audit-md.js',
|
|
63
|
-
// escomplex: same "pattern" parse failure family — BDD scanner
|
|
64
|
-
//
|
|
65
|
-
//
|
|
63
|
+
// escomplex: same "pattern" parse failure family — the BDD scanner walks
|
|
64
|
+
// source trees with regex visitors that hit the upstream destructuring
|
|
65
|
+
// bug.
|
|
66
66
|
'.agents/scripts/lib/bdd-scenario-scanner.js',
|
|
67
|
-
'.agents/scripts/lib/codebase-snapshot.js',
|
|
68
67
|
// escomplex: same "pattern" parse failure — the wave-runner tick uses
|
|
69
68
|
// the regex-property destructuring escomplex chokes on.
|
|
70
69
|
'.agents/scripts/lib/wave-runner/tick.js',
|
|
@@ -1,8 +1,9 @@
|
|
|
1
1
|
/**
|
|
2
2
|
* bdd-scenario-scanner.js — Gherkin scenario index for /plan Phase 7.
|
|
3
3
|
*
|
|
4
|
-
* Story #2637 (sibling to #
|
|
5
|
-
*
|
|
4
|
+
* Story #2637 (sibling to #2636's file-assumption gate; the #2634 and #2635
|
|
5
|
+
* planner-grounding siblings were retired in Story #4811). The Acceptance
|
|
6
|
+
* Engineer step of
|
|
6
7
|
* `epic-plan-spec-author` currently writes ACs from Epic/Tech Spec narrative
|
|
7
8
|
* alone — it never inspects the consumer project's existing `.feature`
|
|
8
9
|
* files. Planned ACs frequently duplicate scenarios that already exist or
|
|
@@ -114,14 +114,6 @@ const KEY_MEANINGS = Object.freeze({
|
|
|
114
114
|
'Allowlist of events that fire a webhook notification.',
|
|
115
115
|
|
|
116
116
|
// planning.*
|
|
117
|
-
'planning.codebaseSnapshot.tier':
|
|
118
|
-
'Depth of the structural codebase view threaded into spec authoring.',
|
|
119
|
-
'planning.codebaseSnapshot.include':
|
|
120
|
-
'Glob patterns included in the codebase snapshot.',
|
|
121
|
-
'planning.codebaseSnapshot.exclude':
|
|
122
|
-
'Glob patterns excluded from the codebase snapshot.',
|
|
123
|
-
'planning.codebaseSnapshot.recentCommitWindow':
|
|
124
|
-
'How many recent commits the snapshot summarizes.',
|
|
125
117
|
'planning.riskHeuristics':
|
|
126
118
|
'Phrases that flag a Story as high-risk for HITL escalation.',
|
|
127
119
|
'planning.failOnSharedEditors':
|
|
@@ -56,6 +56,8 @@ import { mkdtempSync } from 'node:fs';
|
|
|
56
56
|
import os from 'node:os';
|
|
57
57
|
import path from 'node:path';
|
|
58
58
|
|
|
59
|
+
import { reapOnExit } from '../test-temp.js';
|
|
60
|
+
|
|
59
61
|
/**
|
|
60
62
|
* Cache the resolved main-checkout root per spawn cwd so the
|
|
61
63
|
* `git rev-parse` shell-out runs at most once per distinct working
|
|
@@ -223,7 +225,7 @@ function inNodeTestContext(env, execArgv) {
|
|
|
223
225
|
*
|
|
224
226
|
* @param {string} tempRoot
|
|
225
227
|
* @param {NodeJS.ProcessEnv} [env=process.env]
|
|
226
|
-
* @param {{ mkdtemp?: typeof mkdtempSync, execArgv?: string[] }} [deps]
|
|
228
|
+
* @param {{ mkdtemp?: typeof mkdtempSync, execArgv?: string[], onExit?: (fn: () => void) => void }} [deps]
|
|
227
229
|
* Injectable for tests.
|
|
228
230
|
* @returns {string}
|
|
229
231
|
*/
|
|
@@ -238,9 +240,18 @@ export function anchorTempRoot(tempRoot, env = process.env, deps = {}) {
|
|
|
238
240
|
) {
|
|
239
241
|
if (_testContextScratchDir === null) {
|
|
240
242
|
const mkdtemp = deps.mkdtemp ?? mkdtempSync;
|
|
243
|
+
// test-temp-allow: published to children below, so it must live
|
|
244
|
+
// outside the per-process suite root that this process reaps.
|
|
241
245
|
_testContextScratchDir = mkdtemp(
|
|
242
246
|
path.join(os.tmpdir(), 'mandrel-test-temp-'),
|
|
243
247
|
);
|
|
248
|
+
// Creator-only reaping (Story #4808): a process that read the root
|
|
249
|
+
// from the env returned at `scratch` above and never reaches here,
|
|
250
|
+
// so it can never remove a root its parent is still writing to.
|
|
251
|
+
reapOnExit(
|
|
252
|
+
_testContextScratchDir,
|
|
253
|
+
deps.onExit ? { onExit: deps.onExit } : {},
|
|
254
|
+
);
|
|
244
255
|
if (env === process.env) {
|
|
245
256
|
// Children spawned by this test process inherit the same scratch.
|
|
246
257
|
process.env[TEST_TEMP_ROOT_ENV] = _testContextScratchDir;
|
|
@@ -249,35 +249,19 @@ const GITHUB_SCHEMA = {
|
|
|
249
249
|
// rejected as an additional property, so a resurrected key fails loudly rather
|
|
250
250
|
// than silently doing nothing.
|
|
251
251
|
|
|
252
|
-
|
|
253
|
-
|
|
254
|
-
|
|
255
|
-
|
|
256
|
-
|
|
257
|
-
|
|
258
|
-
|
|
259
|
-
|
|
260
|
-
type: 'object',
|
|
261
|
-
properties: {
|
|
262
|
-
tier: { type: 'string', enum: ['skinny', 'medium'] },
|
|
263
|
-
include: {
|
|
264
|
-
type: 'array',
|
|
265
|
-
items: { type: 'string', minLength: 1 },
|
|
266
|
-
},
|
|
267
|
-
exclude: {
|
|
268
|
-
type: 'array',
|
|
269
|
-
items: { type: 'string', minLength: 1 },
|
|
270
|
-
},
|
|
271
|
-
recentCommitWindow: { type: 'integer', minimum: 1 },
|
|
272
|
-
},
|
|
273
|
-
additionalProperties: false,
|
|
274
|
-
};
|
|
252
|
+
// Story #4811: the `planning` block's structural-snapshot key was retired
|
|
253
|
+
// along with the snapshot itself. The pre-computed view it configured grounded
|
|
254
|
+
// nothing — its default include globs missed the standard monorepo layout, and
|
|
255
|
+
// its knobs only re-filtered the same matched set. Spec authoring is grounded
|
|
256
|
+
// by the author's own targeted repo retrieval plus the Phase 8
|
|
257
|
+
// `validateStoryFileAssumptions` gate, neither of which is configurable here.
|
|
258
|
+
// `planning` carries `additionalProperties: false`, so a resurrected key fails
|
|
259
|
+
// loudly; the 2.20.0 retirement migration strips it on upgrade.
|
|
275
260
|
|
|
276
261
|
const PLANNING_SCHEMA = {
|
|
277
262
|
type: 'object',
|
|
278
263
|
properties: {
|
|
279
264
|
riskHeuristics: LIST_OR_EXTENDER_OF_STRINGS,
|
|
280
|
-
codebaseSnapshot: CODEBASE_SNAPSHOT_SCHEMA,
|
|
281
265
|
// Story #4722 (superseding #4683's word-count gate) — shape-derived
|
|
282
266
|
// ceremony-lite routing. Complexity routes on the objective shape of the
|
|
283
267
|
// authored work (changes[] count, acceptance count, creates-vs-refactors
|
|
@@ -60,8 +60,10 @@ import { isObjectPathEntry } from './task-body-validator.js';
|
|
|
60
60
|
/**
|
|
61
61
|
* Default git probe — returns `true` when `path` exists at
|
|
62
62
|
* `baseBranchRef`. Mirrors the existence check used by
|
|
63
|
-
* {@link ./ticket-validator.js#validateAcFreshness}
|
|
64
|
-
*
|
|
63
|
+
* {@link ./ticket-validator.js#validateAcFreshness} so both gates share
|
|
64
|
+
* semantics. (Story #4811 deleted the third sharer, `spec-freshness.js`,
|
|
65
|
+
* along with the codebase snapshot it grounded; this gate — Phase 8 — is
|
|
66
|
+
* now the grounding gate and its behaviour is unchanged.)
|
|
65
67
|
*
|
|
66
68
|
* @param {{ baseBranchRef: string, path: string, cwd?: string }} opts
|
|
67
69
|
* @returns {boolean}
|
|
@@ -48,15 +48,16 @@ const SOURCE_TICKET_FETCH_CONCURRENCY = 4;
|
|
|
48
48
|
* body and ship the raw seed on `seed.content` instead — the budget bounded
|
|
49
49
|
* a field that never left the function.
|
|
50
50
|
*
|
|
51
|
-
*
|
|
52
|
-
* (~
|
|
53
|
-
*
|
|
54
|
-
*
|
|
55
|
-
*
|
|
56
|
-
*
|
|
57
|
-
*
|
|
58
|
-
* under the session budget. The test suite asserts
|
|
59
|
-
* under this value — raise it only with a measured
|
|
51
|
+
* A measured seed-mode envelope on this repo is ~120 KB, dominated by the
|
|
52
|
+
* digest-first `docsContext` (~63 KB inline digest) and the rendered
|
|
53
|
+
* `systemPrompts` (~54 KB); every other field is under 1 KB. Story #4811
|
|
54
|
+
* retired the tier-capped codebase snapshot that used to sit alongside them
|
|
55
|
+
* (~35 KB skinny here). The seed itself is operator-supplied, carried
|
|
56
|
+
* verbatim, and is the only unbounded contributor. 256 KB (~64K tokens at the
|
|
57
|
+
* ≈4-chars/token estimate) leaves roughly 2× headroom over that measurement
|
|
58
|
+
* while staying well under the session budget. The test suite asserts
|
|
59
|
+
* serialized envelopes stay under this value — raise it only with a measured
|
|
60
|
+
* justification.
|
|
60
61
|
*/
|
|
61
62
|
export const PLAN_CONTEXT_ENVELOPE_BYTE_CEILING = 256_000;
|
|
62
63
|
|
|
@@ -112,9 +113,9 @@ function assertPlanContextWithinCeiling(envelope, opts = {}) {
|
|
|
112
113
|
`[plan-context] the assembled "${envelope?.mode}" envelope is ` +
|
|
113
114
|
`${Math.round(bytes / 1024)} KB, over the ` +
|
|
114
115
|
`${Math.round(ceiling / 1024)} KB planner-context ceiling. Largest ` +
|
|
115
|
-
`fields: ${largest}. Trim the seed, plan fewer --tickets source
|
|
116
|
-
'in one run
|
|
117
|
-
'
|
|
116
|
+
`fields: ${largest}. Trim the seed, or plan fewer --tickets source ` +
|
|
117
|
+
'issues in one run. Raising the ceiling needs a measured ' +
|
|
118
|
+
'justification — see PLAN_CONTEXT_ENVELOPE_BYTE_CEILING.',
|
|
118
119
|
);
|
|
119
120
|
}
|
|
120
121
|
|
|
@@ -890,7 +891,6 @@ async function buildSeedFileModeEnvelope({
|
|
|
890
891
|
),
|
|
891
892
|
duplicates,
|
|
892
893
|
docsContext,
|
|
893
|
-
codebaseSnapshot: authoring.codebaseSnapshot,
|
|
894
894
|
bddRunner: authoring.bddRunner,
|
|
895
895
|
bddScenarios: authoring.bddScenarios,
|
|
896
896
|
memoryFreshness: authoring.memoryFreshness,
|
|
@@ -1052,7 +1052,6 @@ async function buildTicketsModeEnvelope({
|
|
|
1052
1052
|
),
|
|
1053
1053
|
duplicates,
|
|
1054
1054
|
docsContext,
|
|
1055
|
-
codebaseSnapshot: authoring.codebaseSnapshot,
|
|
1056
1055
|
bddRunner: authoring.bddRunner,
|
|
1057
1056
|
bddScenarios: authoring.bddScenarios,
|
|
1058
1057
|
memoryFreshness: authoring.memoryFreshness,
|