@elvis1513/auto-coding-skill 4.3.7 → 5.0.0

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 (33) hide show
  1. package/README.md +26 -761
  2. package/cli/assets/managed-install.json +38 -1
  3. package/cli/assets/skill/SKILL.md +15 -100
  4. package/cli/assets/skill/data/templates/ENVIRONMENT.md +30 -0
  5. package/cli/assets/skill/data/templates/architecture/.gitkeep +1 -0
  6. package/cli/assets/skill/data/templates/bridges/AGENTS.md +39 -145
  7. package/cli/assets/skill/data/templates/deployment/.gitkeep +1 -0
  8. package/cli/assets/skill/data/templates/design/.gitkeep +1 -0
  9. package/cli/assets/skill/data/templates/interfaces/.gitkeep +1 -0
  10. package/cli/assets/skill/data/templates/product/.gitkeep +1 -0
  11. package/cli/src/index.js +127 -2589
  12. package/cli/src/sync-assets.js +38 -95
  13. package/package.json +9 -17
  14. package/cli/assets/agents/browser-debugger.toml +0 -27
  15. package/cli/assets/agents/docs-researcher.toml +0 -18
  16. package/cli/assets/agents/explorer.toml +0 -26
  17. package/cli/assets/agents/fixer.toml +0 -29
  18. package/cli/assets/agents/reviewer.toml +0 -50
  19. package/cli/assets/skill/data/contracts/orchestration-v1.schema.json +0 -567
  20. package/cli/assets/skill/data/policies/feedback-resolutions-v1.json +0 -68
  21. package/cli/assets/skill/data/policies/workflow-migrations-v1.json +0 -657
  22. package/cli/assets/skill/data/templates/ENGINEERING.md +0 -115
  23. package/cli/assets/skill/data/templates/tools/ap.py +0 -16
  24. package/cli/assets/skill/requirements.txt +0 -1
  25. package/cli/assets/skill/scripts/ap.py +0 -17791
  26. package/cli/assets/skill/scripts/core.py +0 -510
  27. package/cli/assets/skill/scripts/http_checks.py +0 -92
  28. package/cli/assets/skill/scripts/install_integrity.py +0 -427
  29. package/cli/assets/skill/scripts/scaffold_templates.py +0 -325
  30. package/cli/tests/ap_concurrency_test.py +0 -3602
  31. package/cli/tests/ap_profiles_test.py +0 -3272
  32. package/cli/tests/core_shell_test.py +0 -345
  33. package/cli/tests/index.test.js +0 -2967
package/README.md CHANGED
@@ -1,782 +1,47 @@
1
1
  # auto-coding-skill
2
2
 
3
- A delivery-first Codex engineering workflow:
3
+ `auto-coding-skill` 5.x provides detailed general engineering guidance plus a
4
+ small, project-owned documentation layout. It is not a workflow engine.
4
5
 
5
- `analysis → decomposition → necessary design → development → one final fast
6
- changed-scope gate → commit/push`.
7
-
8
- The Skill is a selectable guardrail, not a command sequence that must run for
9
- every task. The model skips machinery whose expected benefit does not exceed its
10
- cost; read-only work and obvious small clean-checkout changes normally stay direct.
11
-
12
- Version 4.3.7 binds the supervisor's effective retry deadline into the Reviewer
13
- prompt and repairs the exact 4.3.6 stale-deadline block once. Version 4.3.6 accepts the exact equivalent artifact-access wording observed in
14
- real 4.3.4 Reviewer evidence without broadening retry eligibility. Version 4.3.5 keeps retry artifact reads mutation-free and can repair the exact
15
- untouched or procedurally blocked 4.3.4 retry once without rewriting evidence.
16
- Version 4.3.4 adds one fail-closed, immutable-evidence retry for tasks whose
17
- 4.2.8-4.3.2 Reviewer was consumed solely by the fixed Git-local artifact access
18
- defect. Version 4.3.3 resolves the first multi-project feedback batch: read-only
19
- Reviewer artifact access is mutation-free, substantive nonzero-exit results are
20
- preserved, lifecycle identities fail early, managed protocol classification and
21
- database migration signals are precise, and empty feedback status reports the
22
- installed version. Version 4.3.2 keeps upgrade history outside active `docs/`, so an upgrade that
23
- archives the previous managed AGENTS or ENGINEERING contract is immediately
24
- idempotent and `status` remains current. Version 4.3.1 preserves recursively nested project documentation assets of any
25
- file type, and gives project-owned Skill feedback a release-aware lifecycle with
26
- active/closed grouping, recheck and verification notices, and explicit routing
27
- of project preferences into the project overlay. Version 4.3.0 separates exact managed defaults from a byte-stable project
28
- configuration overlay, migrates legacy specialization with semantic proof, and
29
- adds project-owned shared-Skill feedback inboxes with bounded read-only
30
- multi-project triage. Upgrades now use recoverable, owner-bound transactions.
31
- Version 4.2.8 makes the supervised Reviewer observable and recoverable without
32
- weakening review semantics: it streams private event metadata, distinguishes
33
- startup failure from analysis timeout, retries one no-event startup, and supports
34
- an audited fingerprint-bound user override that is never reported as approval.
35
- Version 4.2.7 freezes every pre-commit Reviewer diff as a private, SHA-256-bound
36
- Git-local patch, including staged, unstaged, untracked, deleted, mode, symlink,
37
- and binary changes. Version 4.2.6 makes the managed structure/optimization description defer to
38
- preserved project frontmatter instead of restating template defaults. Version
39
- 4.2.5 keeps JSX self-closing and closing-tag slashes out of regex state.
40
- Version 4.2.4 fixes TypeScript function-body detection after an explicit return
41
- type. Version 4.2.3 restores an explicit opt-in strict size-warning policy without
42
- changing the default, makes semicolon-terminated Java/Kotlin imports visible to
43
- project layer rules, and hardens function-range detection. Version 4.2.2 preserves
44
- project-owned risk and structure policy during upgrades, restores explicit full
45
- and final changed-scope structure enforcement, makes Reviewer assignments
46
- deadline-bound and directly executable, fixes task merge status semantics, and
47
- updates Jenkins/target diagnostics for the current
48
- `access.*` schema while retaining legacy fallback. Version 4.2.1 makes
49
- dirty/concurrent classification fail closed, bounds focused
50
- review latency, provides a complete Reviewer result template, preserves the
51
- caller's runtime environment for routed commands, keeps non-authoritative Agent
52
- history out of structure findings, and distinguishes UI paths and mechanical
53
- renames from semantic changes. Version 4.2.0 makes registered closure resilient
54
- to normal Git staging, reuses an exact successful final gate, and lets an active
55
- task expand scope safely before its first commit.
56
- Version 4.1.9 makes `autocoding init` verify the bounded managed installation after every install or upgrade:
57
- it replaces every managed constraint, migrates only schema-approved project
58
- configuration, and converges `docs/` to one canonical directory framework. Version 4.1.7 made
59
- the Python runtime check, release verification, and tag publication environment
60
- reproducible while removing the unnecessary `requests` dependency. Version 4.1.4
61
- made direct continuation provably pre-write, closed self-review and
62
- risk-classification gaps, validated Agent contracts, and made tag publication
63
- idempotent. Version 4.1.3 closed classification bypasses. Version 4.1.2 made mechanisms required,
64
- model-selectable when beneficial, or forbidden for the current task. Version
65
- 4.1.1 consolidated the 4.x delivery-first workflow into one canonical
66
- repository contract. Version 4.0 replaced the 3.x governance-first defaults with progressive
67
- guardrails. Clean single-writer work stays on the current branch. Worktrees,
68
- parallel fixers, fingerprint review, durable design records, and stronger
69
- affected-scope checks are enabled only when concurrency or risk justifies them.
6
+ It does not require a Gate, classification, task lifecycle, worktree, Reviewer,
7
+ test route, design record, Jenkins check, or deployment check.
70
8
 
71
9
  ## Install
72
10
 
73
11
  ```bash
74
12
  npm install -g @elvis1513/auto-coding-skill
75
- python3 -m pip install PyYAML==6.0.3
76
13
  autocoding init
77
14
  ```
78
15
 
79
- The project install contains:
80
-
81
- ```text
82
- .agents/skills/auto-coding-skill/
83
- .agents/agents/
84
- docs/ENGINEERING.md
85
- docs/project/auto-coding-skill.yaml
86
- docs/tools/autopipeline/ap.py
87
- AGENTS.md (fully managed canonical repository contract)
88
- ```
89
-
90
- `autocoding init` also installs the exact shared documentation tree under
91
- `docs/{architecture,bugs,deployment,design,interfaces,project,reviews,skill-feedback,testing}`.
92
- Managed templates are replaced; project-owned files under the designated
93
- architecture, bugs, deployment, design, interfaces, project, reviews, and testing
94
- roots are preserved recursively and byte-for-byte regardless of extension, as
95
- are `docs/skill-feedback/reports/*.md`;
96
- unrelated
97
- directories are archived under `.agents/archive/`. Re-running init is the complete upgrade operation;
98
- `sync` is the explicit multi-project equivalent. Project-local
99
- `ap.py upgrade --dry-run` remains a read-only legacy diagnostic; its old write
100
- mode is retired so upgrades cannot bypass the transactional installer.
101
- Multi-project sync completes all read-only preflights before its first write;
102
- after that boundary, each project has its own recoverable transaction, so a
103
- later host I/O failure does not roll back projects that already completed.
104
-
105
- `docs/ENGINEERING.md` is the exact managed default layer. Project-specific
106
- access, commands, validation routes, risk rules, structure policy, and other
107
- supported values belong in `docs/project/auto-coding-skill.yaml`. Runtime commands
108
- recursively overlay project mappings on managed defaults; project scalar and list
109
- values replace defaults, while code-enforced safety invariants remain mandatory.
110
- Ordinary init, sync, and upgrade operations never rewrite an existing overlay.
111
- The first compatible upgrade extracts the semantic difference from the previous
112
- managed default, verifies equivalence, and creates the overlay before replacing
113
- managed files.
114
-
115
- ## Shared Skill feedback
116
-
117
- Every initialized project receives:
16
+ Initialization installs a managed general `AGENTS.md` and this lightweight
17
+ documentation layout:
118
18
 
119
19
  ```text
120
- docs/skill-feedback/
121
- ├── README.md
122
- ├── _TEMPLATE-SKILL-FEEDBACK.md
123
- └── reports/ # created when the first report is added
124
- └── YYYY-MM-DD-<slug>-<8-hex>.md
125
- ```
126
-
127
- The README and template are managed; reports are project-owned and upgrades
128
- preserve them byte-for-byte. Releases carry a signature-keyed resolution catalog;
129
- the read-only collector derives upgrade, recheck, verification, closure, and
130
- project-overlay routing actions without rewriting reports. Closed reports are
131
- excluded from active cross-project grouping. A report is a candidate observation only. Create one
132
- after distinguishing managed Skill behavior from project `risk.rules`, validation
133
- routes, access values, structure policy, business code, and environment failures.
134
- Do not automatically turn test failures or Reviewer findings into reports, and do
135
- not include credentials, customer/device data, complete patches, Reviewer
136
- artifacts, raw logs, or absolute user paths.
137
-
138
- Periodic triage is explicit and read-only:
139
-
140
- ```bash
141
- autocoding feedback --projects /path/geestock,/path/geesight,/path/xjmate --json
142
- ```
143
-
144
- The collector scans only bounded report metadata from the listed projects,
145
- executes nothing from the documents, changes no repository, and groups exact
146
- root-cause signatures. Human triage still decides whether an observation is a
147
- shared defect, project configuration, environment issue, duplicate, or missing
148
- evidence before fixes and a release are planned.
149
-
150
- After a project upgrade, verify each requested action. Fixed reports should be
151
- deleted or updated to `resolved`; current regressions should update their last
152
- verified version; project-only preferences belong in
153
- `docs/project/auto-coding-skill.yaml` and should be removed from active feedback.
154
-
155
- Fill the project/Jenkins/GitLab/Nexus URL, username, and password overrides under
156
- `access.*` in `docs/project/auto-coding-skill.yaml`, then configure one real fast
157
- validation command and run:
158
-
159
- ```bash
160
- python3 docs/tools/autopipeline/ap.py doctor
161
- ```
162
-
163
- Plaintext credential values are allowed by the generic workflow.
164
-
165
- ## Adaptive development
166
-
167
- Inspect a task before writing:
168
-
169
- ```bash
170
- python3 docs/tools/autopipeline/ap.py classify --scope auto \
171
- --planned-path backend/internal/orders/service.go \
172
- --intent "fix order retry" \
173
- --writers 1
174
- ```
175
-
176
- The result includes:
177
-
178
- - `execution_mode=direct`: clean checkout and one writer; stay on the current branch.
179
- - `execution_mode=isolated`: dirty checkout, explicit isolation, or multiple writers.
180
- - `execution_mode=none`: no task signal or change; create no branch/worktree.
181
- - `repo`, `workspace_dirty`, `dirty_paths`, and `active_writer`: the repository
182
- and concurrency evidence used for the execution decision.
183
- - `review_required` and `design_required`: risk-based escalation decisions.
184
- - `change_nature`: `mechanical`, `semantic`, or conservative `unknown`; a
185
- mechanical rename may skip a generic durable-design requirement but never a
186
- project rule or required high-risk review.
187
- - `task_kind`: `read_only`, `change`, `terminal_maintenance`, or internal `none`.
188
- - `mechanism_plan.required`: the complete minimum mechanism set for the task.
189
- - `mechanism_plan.optional_when_beneficial`: model-selectable mechanisms whose
190
- expected benefit must exceed coordination cost.
191
- - `mechanism_plan.forbidden`: mechanisms that stay off unless the user overrides.
192
- - `optional_agents`: model-selectable explorer/docs/browser candidates; they are
193
- never automatic stages.
194
- - an adaptive agent plan that does not force fan-out for ordinary work.
195
-
196
- Classification reads the working tree with one fail-closed `git status
197
- --porcelain=v2 -z --untracked-files=all --no-renames` snapshot. A Git error or
198
- malformed result blocks classification instead of being treated as clean. Both
199
- `direct` and dirty/isolated decisions are point-in-time verdicts: reclassify if
200
- the checkout or writer state changes before the first write. Any other active
201
- direct claim is a writer, including another claim with the same owner identity.
202
-
203
- Clean serial work proceeds directly on the current branch with normal Git; it does
204
- not create machine lifecycle state. Use `task-start` only when classification
205
- requires isolation/review or the user explicitly requests lifecycle tracking:
206
-
207
- If a registered task discovers another path before its first commit, expand it
208
- with `task-scope-add` instead of finishing and restarting. Expansion is explicit,
209
- lease/conflict checked, risk-monotonic, and invalidates old review/gate state.
210
- Unregistered direct work should reclassify before writing outside its clean claim.
211
-
212
- ```bash
213
- python3 docs/tools/autopipeline/ap.py task-start T0001 \
214
- --owned-path backend/internal/orders --force-lifecycle
215
- python3 docs/tools/autopipeline/ap.py task-scope-add T0001 \
216
- --owned-path config/orders
217
- python3 docs/tools/autopipeline/ap.py commit-push T0001 \
218
- --msg "T0001: fix order retry"
20
+ docs/ENVIRONMENT.md
21
+ docs/architecture/
22
+ docs/design/
23
+ docs/interfaces/
24
+ docs/deployment/
25
+ docs/product/
219
26
  ```
220
27
 
221
- When a required review still uses the current clean checkout, `task-start` records
222
- a direct manifest without creating a branch/worktree. If no diff is produced,
223
- closure clears the manifest without commit or push.
224
-
225
- Dirty or parallel work receives an isolated task branch/worktree:
226
-
227
- ```bash
228
- python3 docs/tools/autopipeline/ap.py task-start T0002 \
229
- --owned-path backend --review-required
230
- # implement in the printed worktree
231
- python3 docs/tools/autopipeline/ap.py review-run T0002 \
232
- --reviewer "$REVIEWER_ID" --json
233
- python3 docs/tools/autopipeline/ap.py commit-push T0002 \
234
- --msg "T0002: bounded change"
235
- python3 docs/tools/autopipeline/ap.py task-integrate T0002
236
- ```
237
-
238
- Integration serializes the target push and removes clean merged worktrees and
239
- temporary branches. Unknown changes, unmerged history, unsafe ignored data, or
240
- dirty submodules block cleanup rather than being discarded.
241
-
242
- Parallel development uses one task ID and one `task-start` per writer. `--writers`
243
- belongs to `classify` planning only; the runtime rejects overlapping active
244
- `owned_paths` and never creates multiple writers in one task/worktree.
245
-
246
- ## Real changed-scope validation
247
-
248
- Risk classification and validation execution are deliberately separate:
28
+ The environment file is intentionally project-owned and may contain project-
29
+ approved plaintext usernames and passwords for local development and diagnosis.
30
+ Do not invent, echo, or copy credentials beyond the project.
249
31
 
250
- ```yaml
251
- commands:
252
- backend_fast: "cd backend && go test ./internal/orders/..."
253
- frontend_fast: "cd frontend && npm run test:changed"
32
+ All documents are created only when missing and are never overwritten by later
33
+ `init` or `sync` runs. Documentation is optional: record durable facts only when
34
+ the user asks or when they will help future work. Do not create a document for
35
+ every task, review, test, or commit.
254
36
 
255
- validation:
256
- on_unmapped: error
257
- max_command_seconds: 120
258
- max_total_seconds: 180
259
- routes:
260
- - name: backend
261
- paths: ["backend/**", "contracts/**"]
262
- commands: [backend_fast]
263
- timeout_seconds: 90
264
- - name: frontend
265
- paths: ["frontend/**"]
266
- commands: [frontend_fast]
267
-
268
- risk:
269
- rules:
270
- - name: auth
271
- paths: ["backend/**/auth/**"]
272
- profile: high-risk
273
- review: required
274
- ```
275
-
276
- Every path may match multiple routes. All referenced commands are collected in
277
- configuration order, de-duplicated, and executed once. Unmapped code and missing
278
- commands fail before staging. Documentation-only changes may use the built-in
279
- diff check. `git diff --check` remains an additional hygiene check and is never
280
- treated as business validation.
281
-
282
- Route commands run in a non-login Bash that inherits the invoking process PATH
283
- and selected runtime. Projects that intentionally require profile activation
284
- declare it inside the configured command.
285
-
286
- Check coverage without running project commands:
37
+ ## Commands
287
38
 
288
39
  ```bash
289
- python3 docs/tools/autopipeline/ap.py validation-map-check --path contracts/api.yaml
290
- python3 docs/tools/autopipeline/ap.py validation-map-check --tracked
291
- ```
292
-
293
- Focused tests may be run and rerun during implementation. Only the final routed
294
- closure gate is limited to one stable-diff run. Full regression, Docker, builds,
295
- Jenkins, deployment, API verification, and target checks remain explicit
296
- diagnostics outside normal closure.
297
-
298
- Registered tasks call `commit-push` directly because it performs or strictly
299
- reuses that final gate. Run `light-gate` manually only for unregistered normal Git
300
- closure or when explicitly requesting a fresh diagnostic pass. When project
301
- `structure.enabled` is true, the same final gate also runs the changed-scope
302
- structure check; `structure.enforcement: blocking` makes new violations fail.
303
- File/function size thresholds remain warnings by default. They are promoted to
304
- blocking findings only when a project sets both `structure.enforcement: blocking`
305
- and `structure.block_warnings: true`; the template default is `false`, so existing
306
- project behavior does not change merely by upgrading.
307
-
308
- The recommended final closure defaults are 120 seconds per command and 180
309
- seconds total. Projects may raise them for measured affected-scope checks or set
310
- a smaller `timeout_seconds` on a route. A timeout calls for a narrower route and
311
- never triggers a broader fallback.
312
-
313
- ## Risk and subagent policy
314
-
315
- - Micro and standard work default to the main agent.
316
- - Read-only explorer/docs/browser roles run only for independent questions with
317
- clear value.
318
- - Parallel fixers require separate worktrees and non-overlapping paths.
319
- - Reviewer fingerprint approval is required for high-risk, cross-module,
320
- explicitly configured, or parallel integration work.
321
- - Focused review uses `high` effort and a 150-second hard wall over the supplied
322
- stable diff and evidence. Deep review uses `xhigh` and 360 seconds.
323
- - Cross-module or parallel work and API, auth, database, payment, file-transfer,
324
- gateway, or production-configuration boundaries use deep review.
325
- Fixing Reviewer JSON formatting reuses the same analysis and never starts
326
- another substantive review.
327
- - `review-assignment` writes the validated assignment under Git-local state with
328
- the exact fingerprint, HEAD, scope revision, owning fixer, review depth, and
329
- 150/360-second deadline. Before that deadline begins, it freezes the exact
330
- task-owned working-tree state into a mode-0600 binary patch and binds its
331
- canonical path, format, and SHA-256 into the assignment. Reissuing the same
332
- live assignment is idempotent. Snapshot inputs and emitted patches above
333
- 64 MiB fail closed and require a narrower scope or a project-specific
334
- large-artifact review path;
335
- expired or completed attempts cannot be silently renewed, and late results are
336
- rejected by `task-review`.
337
- - `review-run` is the default executable path: it creates/reuses that assignment,
338
- starts a separate `codex exec` session in read-only mode, removes the parent
339
- lifecycle identity, consumes JSON events incrementally, and supervises the full
340
- process group against the original remaining deadline. If no semantic event
341
- arrives within 30 seconds it retries the same immutable assignment once. After
342
- analysis starts it never retries. Private schema-2 receipts record only bounded,
343
- allowlisted event/diagnostic categories, byte counts, and hashes. Plain `review-assignment` remains
344
- available only when another host can enforce the deadline itself; it does not
345
- stop an in-app subagent. A descendant that deliberately creates a new POSIX
346
- session is outside portable process-group termination, but cannot keep the
347
- non-blocking event readers or Reviewer deadline open.
348
- - Exhausted `runtime-unavailable` or `analysis-timed-out` states remain blocked by
349
- default. `review-runtime-override` can record explicit user authorization bound
350
- to the exact fingerprint, assignment, artifact, and runtime receipt. Its verdict
351
- is `runtime-bypassed`, never `approved`; real blocked/changes-requested results
352
- cannot use this path.
353
- - DD/ADR is created only for lasting cross-module, API, data, security,
354
- deployment, or key user-flow decisions.
355
- - Historical debt does not block product work unless the current change worsens it.
356
-
357
- Before returning a Reviewer result, use the absolute assignment path printed by
358
- `review-assignment` to verify and read the frozen patch. Never reconstruct a
359
- pre-commit review from `diff_base..diff_head` or a live `git diff`:
360
-
361
- ```bash
362
- python3 docs/tools/autopipeline/ap.py review-artifact \
363
- --file <assignment.json>
364
- ```
365
-
366
- Then generate the complete 16-field skeleton and fill its summary, evidence,
367
- findings, and risks:
368
-
369
- ```bash
370
- python3 docs/tools/autopipeline/ap.py agent-result-template \
371
- --file <assignment.json> --verdict <approved|changes-requested|blocked>
372
- ```
373
-
374
- The generated object prevents field-by-field contract retries; contract checking
375
- also reports all known field errors together. If only its JSON shape needs repair,
376
- correct that object without rerunning review analysis.
377
-
378
- The main agent owns architecture, final validation, Git closure, ordered
379
- integration, push, and cleanup. Push ends the coding task; later CI/acceptance is
380
- not polled automatically. An explicitly requested failure diagnosis may continue
381
- in the same conversation/task without an artificial second ledger lifecycle.
382
-
383
- ## Upgrade projects
384
-
385
- Finish every registered task using its currently installed runtime before changing
386
- the Skill version. Then run `autocoding init` from each project root. It is safe to
387
- rerun and needs no force flag.
388
-
389
- The only migration exception is an active 4.2.8-4.3.2 task whose completed
390
- Reviewer result contains no findings and is blocked solely by the fixed
391
- `review-artifact` Git-local permission error. Run the managed 4.3.7 runtime from
392
- outside the project install to authorize one audited retry, then immediately run
393
- the original Reviewer identity:
394
-
395
- ```bash
396
- python3 /absolute/managed/4.3.7/ap.py --repo /path/project \
397
- review-runtime-retry T0001 --diff-fingerprint <SHA256> \
398
- --reason-code managed-review-artifact-access \
399
- --confirm-managed-runtime-retry
400
- python3 /absolute/managed/4.3.7/ap.py --repo /path/project \
401
- review-run T0001 --reviewer <ORIGINAL_REVIEWER> --json
402
- ```
403
-
404
- The command preserves the original assignment, patch, result, receipt, and event
405
- log byte-for-byte. It refuses substantive findings, changed scope or fingerprint,
406
- tampered evidence, user overrides, duplicate retries, and unmanaged runtimes.
407
- If 4.3.4 already authorized the same retry, a fixed runtime may supersede only an untouched
408
- pending attempt or its exact non-substantive artifact-access block; both audit
409
- chains and any prior runtime files remain immutable. Version 4.3.7 may also
410
- supersede the exact 4.3.6 block that used the immutable assignment's old deadline
411
- after successfully reading the artifact; ordinary timeouts remain terminal.
412
-
413
- ```bash
414
- cd /path/a && autocoding init
415
- cd /path/b && autocoding init
416
- autocoding status --projects /path/a,/path/b
417
- ```
418
-
419
- Init replaces the managed Skill, root `AGENTS.md`, managed agents, ENGINEERING
420
- schema/body, runtime launcher, and documentation framework. It preserves explicit
421
- model overrides, complete project `risk.rules`, supported project/access/
422
- concurrency/route/structure values, and an existing project-owned structure
423
- standard byte-for-byte. Removed content is archived outside active docs.
424
-
425
- ## What changed in 4.3.7
426
-
427
- - Passed the supervisor-authorized effective deadline in both the Reviewer prompt
428
- and environment, explicitly superseding the immutable assignment deadline only
429
- for an audit-bound retry.
430
- - Added one fail-closed repair for the exact 4.3.6 stale-deadline block. It requires
431
- successful artifact verification, no findings or changed paths, an observed time
432
- between the original and retry deadlines, and immutable evidence from every
433
- prior generation.
434
-
435
- ## What changed in 4.3.6
436
-
437
- - Added the exact `review-artifact exited with:` evidence, summary, and risk
438
- wording emitted by a real 4.3.4 Reviewer. Findings, changed paths, timeouts,
439
- overrides, and every unrelated failure remain ineligible for migration.
440
-
441
- ## What changed in 4.3.5
442
-
443
- - Stopped read-only `review-artifact` from recomputing the live task snapshot;
444
- the writable supervisor remains responsible for the immediately pre-launch
445
- HEAD, scope, and fingerprint check.
446
- - Added one fail-closed repair for `retry-v4.3.4` when it is still untouched or
447
- ended only in the known Git-local permission block. The repair publishes a new
448
- audit and tokenized evidence while retaining and revalidating the full 4.3.4
449
- audit chain.
450
-
451
- ## What changed in 4.3.4
452
-
453
- - Added `review-runtime-retry`, a one-attempt migration for the exact
454
- 4.2.8-4.3.2 read-only `review-artifact` permission defect. Eligibility is bound
455
- to the lifecycle owner, original Reviewer, task UUID, base/HEAD/scope,
456
- fingerprint, canonical non-substantive blocked result, and a strictly newer
457
- managed runtime.
458
- - Kept the original assignment, frozen patch, result, runtime receipt, and event
459
- log immutable. The retry writes a mode-0600 create-only audit plus tokenized
460
- result/run/events files and carries its own fixed 150/360-second deadline.
461
- - Made `review-run`, `review-artifact`, `task-review`, approval validation, and
462
- task status validate the retry audit and effective deadline. Artifact access
463
- receives the extension only through the supervised runtime environment and
464
- the exact absolute managed script that authorized it.
465
-
466
- ## What changed in 4.3.3
467
-
468
- - Made existing Git-local Reviewer storage validation read-only: the writable
469
- supervisor establishes mode `0700`, while the Reviewer verifies ownership and
470
- mode without calling `chmod` from its read-only sandbox. Reviewer artifact and
471
- result-template commands now use the exact absolute runtime that launched
472
- `review-run`, allowing a fixed runtime to review a task created by an older
473
- project installation without changing that task's base or owned scope.
474
- - Preserved contract-valid `blocked` and `changes-requested` results when the
475
- Reviewer process exits nonzero, while retaining the blocking runtime receipt.
476
- - Rejected `task-start` owner/writer values that conflict with
477
- `CODEX_THREAD_ID` before creating task state or worktrees.
478
- - Classified root `AGENTS.md` as managed release/tooling authority and prevented
479
- unqualified workflow/configuration migration wording or policy filenames from
480
- being mislabeled as database changes.
481
- - Reported the installed Skill version in status even before a project creates
482
- its first feedback report.
483
- - Added release-catalog resolutions for all nine reports collected from the
484
- seven-project 4.3.2 rollout, including canonical duplicate mappings.
485
-
486
- ## What changed in 4.3.2
487
-
488
- - Moved managed AGENTS and duplicate-workflow ENGINEERING history into
489
- `.agents/archive/auto-coding-skill/<version>/` instead of `docs/archive/`.
490
- - Added an upgrade regression proving legacy 4.3.1 active-doc archives are moved
491
- out of active docs without byte loss and post-upgrade `status` is immediately
492
- current.
493
-
494
- ## What changed in 4.3.1
495
-
496
- - Replaced extension-bound documentation preservation with recursive ownership
497
- for designated project documentation roots. PNG, SVG, JSON, YAML, binary, and
498
- nested assets now remain byte-identical across status, dry-run, init, and sync;
499
- active documentation symlinks and special files fail before writes.
500
- - Added the managed, signature-keyed feedback resolution catalog and compatible
501
- v1/v2 report parsing. Collection now separates active and closed reports and
502
- derives release-aware upgrade, recheck, verification, regression, closure, and
503
- project-overlay routing states without modifying project-owned reports.
504
- - Surfaced feedback maintenance as non-blocking status, sync, init, and doctor
505
- advisories, while keeping explicit feedback collection strict, bounded,
506
- metadata-only, and read-only.
507
-
508
- ## What changed in 4.3.0
509
-
510
- - Made `docs/ENGINEERING.md` the exact managed default layer and added
511
- `docs/project/auto-coding-skill.yaml` as the higher-priority project-owned
512
- overlay. Mappings merge recursively; explicit project scalars and lists replace
513
- defaults, while managed workflow identity and code-enforced safety invariants
514
- remain protected.
515
- - Added one-time semantic migration from the previous manifest-bound default.
516
- Initialization creates the overlay without overwriting an existing one, proves
517
- the reconstructed effective configuration is equivalent, archives the legacy
518
- document, and then converges the managed default byte-for-byte.
519
- - Added `docs/skill-feedback/` with managed guidance/template and preserved
520
- project-owned reports. `autocoding feedback --projects ... --json` performs an
521
- explicit bounded metadata-only read and groups exact root-cause signatures for
522
- later human triage.
523
- - Replaced partial upgrade writes with owner-bound, recoverable installation
524
- transactions, staged/installed hash checks, active-installer rejection, safe
525
- project-file operations, and fail-closed effective-configuration validation.
526
- Project-local `ap.py upgrade --write` is retired in favor of transactional
527
- `autocoding init` or explicit multi-project `sync`.
528
- - Batched protected executable-mode convergence so the safer installer retains
529
- practical upgrade latency while preserving symlink/reparse checks.
530
-
531
- ## What changed in 4.2.8
532
-
533
- - Replaced end-only Reviewer output collection with bounded JSONL event streaming.
534
- Schema-2 Git-local receipts now record CLI/model/effort identity, attempt phases,
535
- first/last event times, event-log hashes, diagnostic categories, and byte counts
536
- without persisting prompt, patch, stderr, tool output, or model content.
537
- - Split runtime startup failure from analysis timeout. A no-semantic-event startup
538
- is capped at 30 seconds and retried once with the same assignment, fingerprint,
539
- artifact, command, and absolute deadline; an analysis that started is never
540
- repeated. Focused review now uses `high/150s`, while deep uses `xhigh/360s`.
541
- - Added `review-runtime-override` for explicit user-authorized delivery after an
542
- exhausted runtime-unavailable or analysis-timeout failure. The private override
543
- binds all immutable review evidence and records `runtime-bypassed`, never
544
- `approved`; failure-time hashes reject pre-authorization tampering, and a
545
- substantive Reviewer result observed before timeout remains non-bypassable.
546
- - Restricted the hidden custom runner to explicit test harnesses and closed spawn
547
- failure, unbounded-output, silent-startup, and lost-diagnostics paths.
548
-
549
- ## What changed in 4.2.7
550
-
551
- - Replaced live pre-commit diff discovery with an immutable Git-local patch
552
- artifact that captures staged, unstaged, untracked, deleted, mode, symlink,
553
- and binary changes in both direct and isolated task modes.
554
- - Bound the artifact's canonical path, format, and SHA-256 into the Reviewer
555
- assignment, then bound the complete assignment SHA-256 into runtime state,
556
- receipt, verdict validation, and final commit authorization. The deadline
557
- starts only after snapshot creation.
558
- - Added the full `review-artifact` command for verified Reviewer access and
559
- lifecycle cleanup for task artifacts after successful direct closure,
560
- isolated cleanup, or explicit orphan pruning. Cleanup now preserves registry
561
- and review evidence on failure so the same task can be retried safely.
562
-
563
- ## What changed in 4.2.6
564
-
565
- - Made the managed `structure` / `optimization` description configuration-neutral:
566
- the preserved project frontmatter is authoritative for enforcement and debt
567
- completion policy.
568
- - Added a 4.2.5-to-4.2.6 initialization regression that preserves strict
569
- `blocking + baseline-aware` project settings and proves repeated init remains
570
- byte-idempotent.
571
-
572
- ## What changed in 4.2.5
573
-
574
- - Prevented standalone JSX `/>` and closing-tag `</...>` slashes from starting a
575
- JavaScript regex token during function-range scans. GeeStack now matches its
576
- TypeScript AST size baseline, and XJMate strict full structure passes cleanly.
577
-
578
- ## What changed in 4.2.4
579
-
580
- - Fixed TypeScript/TSX named and arrow function ranges across generic constraints,
581
- predicates, conditional/object/function return types, and multiple same-line
582
- arrows. Exact arrow locations are evaluated with bounded longest-range selection,
583
- so type arrows neither truncate real bodies nor create duplicate findings.
584
-
585
- ## What changed in 4.2.3
586
-
587
- - Added the explicit `structure.block_warnings` policy with a default of `false`.
588
- File/function size threshold findings become blocking only when the project also
589
- sets `structure.enforcement: blocking`; advisory and default configurations keep
590
- their existing warning-only behavior.
591
- - Recognized semicolon-terminated Java and Kotlin imports during structure checks,
592
- so project-configured forbidden-import layer rules can enforce JVM boundaries.
593
- - Hardened function-range detection for Go receivers, Python suites, brace bodies,
594
- and block/expression arrow functions. Scans are threshold-bounded, honor the
595
- final Gate deadline, and ignore braces in comments and string literals.
596
-
597
- ## What changed in 4.2.2
598
-
599
- - Classified root and nested `package.json` files as `release_or_tooling` high-risk
600
- changes requiring review, and added upgrade regression coverage that preserves
601
- complete project `risk.rules` instead of deleting partially overlapping rules.
602
- - Honored explicit `structure-check --scope full`, included tracked and unignored
603
- untracked files, and restored the enabled structure check inside the final
604
- changed-scope gate. Init now preserves supported strict structure settings and
605
- treats `docs/architecture/structure-standard.md` as scaffold-once project policy.
606
- - Added `review-run` with a validated Git-local assignment, fixed 90/300-second
607
- process-group supervision, one attempt per fingerprint, Reviewer identity/HEAD/
608
- scope binding, safe result normalization, and late-result rejection. The child
609
- runs as a separate read-only Codex process without the lifecycle-owner identity;
610
- the documented template command uses the complete
611
- `python3 docs/tools/autopipeline/ap.py ...` invocation.
612
- - Added `has_task_commits` and prevented a no-commit task from reporting
613
- `merged_into_target=true` merely because its baseline equals the target tip.
614
- - Updated `verify-jenkins`, `verify-jenkins-build`, `verify-target`, and target
615
- health checks to prefer `access.jenkins.<component>` and
616
- `access.project.<component>`, fail before network activity on missing or
617
- ambiguous endpoints, isolate Jenkins crumbs by URL and username, and preserve
618
- legacy `jenkins.*` / `target_env.*` compatibility.
619
-
620
- ## What changed in 4.2.1
621
-
622
- - Made classification take one fail-closed Git porcelain-v2 snapshot and expose
623
- `repo`, `workspace_dirty`, `dirty_paths`, and `active_writer` in normal output.
624
- Direct/dirty decisions are explicit snapshots, and another same-owner direct
625
- claim still counts as a concurrent writer.
626
- - Kept the `reviewer` Agent at `xhigh` while adding one-pass, 90-second focused
627
- review and 300-second deep routing for cross-module, parallel, API, auth,
628
- database, payment, file-transfer, gateway, and production-configuration changes.
629
- Focused timeout is `blocked`, never an approval or a reason to call `task-review`.
630
- - Added `agent-result-template` to generate all 16 Reviewer result fields from a
631
- validated assignment and made contract validation return all known field errors
632
- together. JSON-only repair reuses the completed analysis instead of launching
633
- another substantive review.
634
- - Replaced the login Shell used for routed commands with an inherited-environment
635
- non-login Shell, preserving the caller's PATH and selected Python/Node runtime.
636
- - Excluded `.agents/archive/**`, managed Agent definitions, and the install
637
- manifest from structure scans while continuing to scan project-owned `.agents` code.
638
- - Replaced substring UI classification with exact directory segments and UI-only
639
- extensions, removing false browser signals from files such as `.env.components`.
640
- - Added conservative mechanical/semantic/unknown change classification. Exact
641
- renames and explicitly behavior-stable syncs can avoid generic cross-module
642
- design ceremony; semantic/unknown changes and explicit project design rules
643
- remain protected, and high-risk review is unchanged.
644
-
645
- ## What changed in 4.2.0
646
-
647
- - Made `commit-push` preserve already staged deletions and mixed index/worktree
648
- state, and made ownership/validation see both ends of a rename.
649
- - Added an exact per-worktree final-gate PASS receipt. Registered `commit-push`
650
- reuses it only when content, index, base, scope, validation plan, command hashes,
651
- and writer lease still match; staging or push retries no longer rerun a stable gate.
652
- - Added `task-scope-add` for conflict-checked, monotonic scope expansion before the
653
- first commit. Expansion raises risk when necessary and invalidates review/gate state.
654
- - Clarified that registered tasks call `commit-push` directly; standalone
655
- `light-gate` is for unregistered Git closure or an explicit fresh diagnostic.
656
-
657
- ## What changed in 4.1.9
658
-
659
- - Added a packaged managed-install manifest with path, version, content hash, and
660
- executable-bit expectations for release-owned files.
661
- - Made `autocoding init` verify the written Skill, managed agents, root protocol,
662
- launcher, managed ENGINEERING region, and canonical templates before success.
663
- - Made `doctor` recheck the local manifest and `status` compare it with the current
664
- release while excluding project facts, durable records, custom agents, and archives.
665
-
666
- ## What changed in 4.1.8
667
-
668
- - Made `autocoding init` perform a complete project install or upgrade without a
669
- separate sync/upgrade chain or `--force`.
670
- - Rebuilt ENGINEERING from the current schema, preserving only supported project
671
- values and removing unknown legacy fields and competing workflow text.
672
- - Made the generated docs directory topology and managed templates identical
673
- across projects while preserving mutable project docs and valid durable records;
674
- obsolete paths are archived under `.agents/archive/` before removal.
675
- - Made root AGENTS, managed agents, the Skill copy, and the project launcher fully
676
- converge during init.
677
-
678
- ## What changed in 4.1.7
679
-
680
- - Made tag publication reproducible with explicit Node 24, Python 3.12, and pinned
681
- PyYAML setup before package verification.
682
- - Made `autocoding init` fail before writes with one interpreter-specific recovery
683
- command when its Python runtime cannot import PyYAML.
684
- - Removed the `requests` runtime dependency by using Python's standard HTTP library.
685
- - Made concurrency tests self-contained instead of borrowing local thread and Git
686
- identity configuration.
687
- - Supported both GitHub-secret publication and local-token publication followed by
688
- idempotent tag verification.
689
-
690
- ## What changed in 4.1.4
691
-
692
- - Added clean pre-write direct claims so broad paths cannot adopt unknown existing
693
- dirt after development starts.
694
- - Prevented the current writer lease holder from approving its own diff and added
695
- executable assignment/result contract validation.
696
- - Escalated high-confidence database, authorization, settlement, gateway, and
697
- production intents while keeping ordinary UI work lightweight.
698
- - Made mechanism plans dependency-closed: parallel fixers require reclassification
699
- with multiple writers and therefore require isolated worktrees.
700
- - Published from matching `v*` tags with idempotent npm registry checks and a final
701
- registry verification.
702
-
703
- ## What changed in 4.1.3
704
-
705
- - Made explicit task kinds fail closed: `none` is internal-only and terminal
706
- maintenance cannot be applied to code or unrelated documentation paths.
707
- - Added guarded `--continue-direct` reclassification for already-owned changes;
708
- unknown dirty paths still require isolation.
709
- - Separated intent risk hints from path/rule-confirmed high risk so ordinary login
710
- or payment UI work no longer automatically receives a heavy lifecycle.
711
- - Aligned generated Agent plans with their JSON Schema and exposed optional
712
- read-only Agent candidates without auto-dispatching them.
713
- - Clarified delegated-fixer worktree ownership, reviewer identity, and explorer
714
- source authority.
715
-
716
- ## What changed in 4.1.2
717
-
718
- - Added explicit read-only, change, and terminal-maintenance task kinds so
719
- intent-only questions cannot accidentally trigger code-delivery machinery.
720
- - Split mechanism planning into required, model-selectable, and forbidden sets;
721
- read-only discovery agents remain available when they save time or add needed
722
- expertise.
723
- - Made 120/180-second gate budgets recommended defaults, added route-level
724
- timeouts, and kept project-specific affected-scope overrides valid.
725
- - Rejected unnecessary lifecycle creation before access checks, fetching, locks,
726
- or branch work; corrected the review lifecycle examples.
727
- - Added a release-package assertion and prepack cache cleanup so generated Python
728
- caches cannot leak into the npm tarball.
729
-
730
- ## What changed in 4.1.1
731
-
732
- - Added a machine-readable minimum mechanism plan and rejected unnecessary clean
733
- serial task lifecycles unless explicitly requested.
734
- - Added bounded per-command and total time budgets for the final changed-scope gate.
735
- - Made root `AGENTS.md` the sole behavioral protocol, reduced `SKILL.md` to
736
- invocation guidance, and reduced `ENGINEERING.md` to project configuration and
737
- facts.
738
-
739
- ## What changed in 4.1.0
740
-
741
- - Made root `AGENTS.md` byte-identical across installed projects and whole-file
742
- managed on every full sync/upgrade.
743
- - Established code/tests/runtime as the source of current behavior and
744
- `ENGINEERING.md` as the single project workflow/configuration source.
745
- - Added an engineering-centered documentation framework plus optional
746
- `docs/project/` fact templates.
747
- - Prevented Skill upgrades while any registered task is active and recorded the
748
- installed Skill version in new task manifests.
749
- - Made reviewer scope bounded, with targeted recheck for mechanical docs-only
750
- fixes and non-blocking handling of adjacent findings.
751
- - Defined ledger/archive reconciliation as a terminal maintenance action, so
752
- closing records cannot recursively create more task records.
753
- - Kept explicit Jenkins failure diagnosis in the same requested task while
754
- preserving push-as-completion and no automatic polling.
755
-
756
- ## What changed in 4.0.0
757
-
758
- - Added real `direct` and `isolated` runtime execution modes.
759
- - Stopped creating branches/worktrees for clean serial or no-diff work.
760
- - Added multi-match `validation.routes`, stable command de-duplication,
761
- fail-closed unmapped code, and `validation-map-check`.
762
- - Separated `risk.rules` from executable validation.
763
- - Made ordinary tasks main-agent-first and review/design conditional.
764
- - Kept advanced worktree ownership, dependency SHA, lease, fingerprint,
765
- integration, submodule safety, and branch cleanup for parallel/high-risk work.
766
- - Removed taskbook, closure, evidence, and active-task documents from the default
767
- scaffold and normal closure path.
768
- - Changed structure/debt policy to no-new-debt and kept full/CI/runtime checks as
769
- explicit diagnostics.
770
- - Added schema-3 task manifests and atomic sync refusal for in-flight pre-4.0 tasks.
771
-
772
- ## Maintainer commands
773
-
774
- ```bash
775
- npm run sync-assets
776
- npm run release:check
40
+ autocoding init
41
+ autocoding sync --projects /path/project-a,/path/project-b
42
+ autocoding status --projects . --json
777
43
  ```
778
44
 
779
- `release:check` validates source/assets, Python 3.11 grammar, managed Agent TOML,
780
- CLI/profile/concurrency regressions, and the npm package contents.
781
-
782
- License: MIT.
45
+ These commands only install, preserve, and report documentation. They never run
46
+ project tests, validation, Gates, reviews, builds, Jenkins jobs, deployment, or
47
+ other external-system operations.