@elvis1513/auto-coding-skill 3.0.3 → 4.0.0

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.
@@ -1,5 +1,6 @@
1
1
  ---
2
2
  workflow:
3
+ skill_version: "4.0.0"
3
4
  mode: "dev"
4
5
  profile: "auto"
5
6
  completion: "push"
@@ -49,337 +50,181 @@ concurrency:
49
50
  disposable_ignored: []
50
51
 
51
52
  commands:
52
- gate_changed: "git diff --check"
53
-
54
- gate:
55
- default_scope: "changed"
56
- fallback_scope: "changed"
57
- full_on_unknown: false
58
- no_change_scope: "changed"
59
- profile_log: ".local/auto-coding-skill/gate-profile.jsonl"
53
+ # Configure one real, fast project-native command. `autocoding sync` fills
54
+ # this automatically only when package.json declares `test:changed`.
55
+ project_fast: ""
56
+
57
+ validation:
58
+ on_unmapped: "error"
59
+ routes:
60
+ - name: "project-code"
61
+ paths: ["**"]
62
+ exclude: ["*.md", "docs/**"]
63
+ commands: ["project_fast"]
64
+
65
+ risk:
60
66
  rules: []
61
67
 
62
68
  structure:
63
69
  enabled: true
64
70
  enforcement: "advisory"
65
71
  architecture_standard: "project-defined"
66
- max_file_lines_warn: 800
67
- max_file_lines_block: 1500
68
- max_function_lines_warn: 120
69
- max_added_lines_to_large_file: 80
70
- require_reuse_search: true
71
- block_new_responsibility_in_large_file: true
72
- allow_large_files:
73
- - ".agents/skills/**"
74
- - "docs/tools/autopipeline/**"
75
- - "generated/**"
76
- - "**/generated/**"
77
- - "vendor/**"
78
- - "dist/**"
79
- - "build/**"
80
- - "target/**"
81
- - "node_modules/**"
82
- - "**/*.generated.*"
83
- - "**/*.min.js"
84
- - "**/*.map"
85
72
  accepted_debt_paths: []
86
73
  layer_rules:
87
74
  enabled: true
88
75
  block: false
89
76
 
90
77
  optimization:
91
- completion_policy: "baseline-aware"
78
+ completion_policy: "no-new-debt"
92
79
  require_baseline_for_global_review: false
93
80
  report_accepted_debt_as_findings: false
94
81
 
95
82
  docs:
96
- taskbook: "docs/tasks/taskbook.md"
97
- closure_log: "docs/tasks/closure-log.md"
98
- evidence_log: "docs/tasks/evidence.jsonl"
99
- active_task_dir: "docs/tasks/active"
100
- task_closure_dir: "docs/tasks/closures"
101
- task_evidence_dir: "docs/tasks/evidence"
102
83
  design_dir: "docs/design"
103
- task_archive_dir: "docs/tasks/archives"
104
- design_archive_dir: "docs/archive/design"
105
- archive_index: "docs/tasks/archive-index.md"
106
- ledger_check_enabled: true
107
- ledger_block_on_exceed: true
108
- active_taskbook_max_lines: 1200
109
- active_closure_log_max_lines: 800
110
- active_design_max_files: 120
111
84
  health_baseline: "docs/reviews/project-health-baseline.md"
112
85
  optimization_backlog: "docs/reviews/optimization-backlog.md"
113
86
  structure_standard: "docs/architecture/structure-standard.md"
114
87
  api_docs_required: false
115
88
  api_doc: "docs/interfaces/api.md"
116
89
  api_change_log: "docs/interfaces/api-change-log.md"
117
- regression_matrix: "docs/testing/regression-matrix.md"
118
- bug_list: "docs/bugs/bug-list.md"
119
- summary_dir: "docs/tasks/summaries"
120
90
  ---
121
91
 
122
92
  # Engineering Workflow
123
93
 
124
- <!-- auto-coding-skill:managed-workflow:start version=3.0.3 -->
125
-
126
- This file is the single project workflow configuration. Keep it in Git. Its YAML
127
- frontmatter and content outside these versioned markers are manually maintained;
128
- `autocoding sync` updates only this managed workflow block. Fill every access URL,
129
- username, and password during project initialization.
130
-
131
- ## Execution profiles
132
-
133
- `workflow.profile: auto` selects one of three effective profiles:
134
-
135
- | Profile | Typical work | Local gate | Completion |
136
- | --- | --- | --- | --- |
137
- | `micro` | docs/tests-only, very small isolated work | changed/quick | pushed |
138
- | `standard` | ordinary feature and defect work | changed/quick | pushed |
139
- | `high-risk` | DB/auth/payment/deploy/build config or declared high risk | changed/quick | pushed |
140
-
141
- `auto` is only a selector. The effective profile remains useful for analysis,
142
- design depth, and reviewer recommendations, but it never expands the local gate.
143
- All profiles run the fast changed-scope gate and complete when the configured
144
- target branch is pushed. Jenkins owns later build/deploy work; its result and
145
- manual acceptance are outside the coding task.
146
-
147
- Project `gate.rules` may raise the planning profile or request design/reviewer
148
- attention. Their legacy `scope` and `commands` values never execute in the
149
- automatic development flow. Standard/full commands remain callable on demand
150
- with `ap.py run <name>` but are not part of normal development.
151
-
152
- ## Minimal configuration
153
-
154
- Required for normal development:
155
-
156
- - `workflow.mode`: `dev`
157
- - `workflow.profile`: `auto`, `micro`, `standard`, or `high-risk`
158
- - `workflow.completion`: `push`
159
- - `project.name`
160
- - all URL, username, and direct password fields under `access.project`,
161
- `access.jenkins`, `access.gitlab`, and `access.nexus`; blank/TODO values make
162
- initialization incomplete and block `doctor`; keep each value as an inline
163
- YAML string, quoting values that resemble numbers, dates, booleans, or YAML
164
- collections
165
- - `concurrency.isolation`: use `adaptive`; a clean serial task works directly on
166
- the current branch, while dirty or parallel write work uses isolated worktrees
167
- - `concurrency.base_ref` and `target_branch`: leave blank to derive the current
168
- upstream, or set them explicitly for a fixed integration branch
169
- - `concurrency.branch_prefix` and `worktree_root`
170
- - `concurrency.cleanup_merged` and `delete_remote_branch`: keep both enabled to
171
- remove integrated temporary worktrees and branches
172
- - `concurrency.disposable_ignored`: add only project-specific ignored cache/build
173
- paths that may be discarded with a task worktree; ignored data and unknown
174
- paths block cleanup, including inside initialized submodules
175
- - one fast `commands.gate_changed`; the default is `git diff --check`
176
-
177
- Legacy command keys such as `commands.light_gate`, `test`, and `build` remain
178
- callable explicitly with `ap.py run <name>` but never replace the automatic
179
- changed gate. `quick_test` remains the only fallback when `gate_changed` is
180
- absent. Run `doctor` for exact missing-field diagnostics.
181
-
182
- If the actual changed gate fails only because a dependency already declared in
183
- the repository lockfile is not installed in the working checkout, restore that
184
- locked dependency locally (for example with `npm ci`) and rerun the same changed
185
- gate once. Do not install dependencies or restart a standard/full diagnostic
186
- that was invoked outside the normal development gate, and never promote this
187
- recovery into a full-gate run.
188
-
189
- ## Concurrent write tasks
190
-
191
- For one serial write task, inspect the current checkout first. If it is clean,
192
- develop directly on the current branch, run the changed gate once, commit, and
193
- push without creating a temporary branch. If it already contains changes, or if
194
- multiple writers will run concurrently, preserve the checkout and use one
195
- registered task branch/worktree per writer. Never share one worktree between
196
- writers. Clean temporary worktrees and merged task branches after integration.
94
+ <!-- auto-coding-skill:managed-workflow:start version=4.0.0 -->
95
+
96
+ This file contains project facts and the small amount of configuration the
97
+ generic workflow cannot infer. Normal delivery is:
98
+
99
+ `analysis → decomposition → necessary design → development → one final fast
100
+ changed-scope gate → commit/push`.
101
+
102
+ Target-branch push ends the coding task. Jenkins, deployment, and owner
103
+ acceptance happen later unless the user opens an explicit diagnostic task.
104
+
105
+ ## Delivery levels
106
+
107
+ | Level | Default behavior | Added guardrails |
108
+ | --- | --- | --- |
109
+ | Micro | main agent, current branch, fast route | none |
110
+ | Standard | main agent or justified read-only help | no mandatory reviewer or design document |
111
+ | High risk | focused design and independent review | only the affected validation routes |
112
+ | Parallel | one isolated writer per unit | owned paths, dependency SHAs, leases, fingerprint review, ordered integration |
113
+
114
+ Risk changes analysis, design, and review depth. It never promotes normal local
115
+ development to a full repository build, Docker run, Jenkins poll, deployment, or
116
+ target-environment acceptance.
117
+
118
+ ## Adaptive isolation
119
+
120
+ Run task planning before development:
197
121
 
198
122
  ```bash
199
- python3 docs/tools/autopipeline/ap.py task-start <TASK_ID> \
200
- --owned-path <PATH> [--owned-path <PATH> ...] \
201
- [--depends-on <TASK_ID>=<SHA> ...] [--writer <WRITER>]
202
- # Enter the worktree path printed by task-start.
203
- python3 docs/tools/autopipeline/ap.py task-status <TASK_ID>
204
- python3 docs/tools/autopipeline/ap.py task-submodule-sync <TASK_ID>
205
- python3 docs/tools/autopipeline/ap.py task-review <TASK_ID> \
206
- --verdict approved --diff-fingerprint <SHA256> [--reviewer <REVIEWER>]
207
- python3 docs/tools/autopipeline/ap.py task-handoff <TASK_ID> \
208
- --from <WRITER> --to <WRITER> [--generation <N>]
209
- python3 docs/tools/autopipeline/ap.py commit-push <TASK_ID> \
210
- --msg "<TASK_ID>: <summary>" [--writer <WRITER>]
211
- python3 docs/tools/autopipeline/ap.py task-integrate <TASK_ID>
212
- python3 docs/tools/autopipeline/ap.py task-resume <TASK_ID>
123
+ python3 docs/tools/autopipeline/ap.py classify --scope auto \
124
+ --planned-path <PATH> --intent "<task intent>" [--writers <N>]
213
125
  ```
214
126
 
215
- `task-start` fixes the task's base revision and records its task branch and
216
- worktree, repeated `owned_paths`, dependency task/SHA pairs, and the active writer
217
- lease in the repository manifest. Run `commit-push` only from that registered
218
- worktree. Changed and staged paths must be a subset of `owned_paths`; an unknown
219
- change blocks the push and must never be restored, reset, stashed, cleaned, or
220
- otherwise altered.
221
-
222
- The lifecycle fields are runtime gates, not advisory notes. `task-review` records
223
- the reviewer verdict against the current diff fingerprint; any edit, rebase, or
224
- conflict resolution invalidates that approval. `task-handoff` is the only way to
225
- transfer the writer lease, and its generation check prevents stale owners from
226
- resuming writes. `commit-push` validates the active writer (explicit `--writer`
227
- or `CODEX_THREAD_ID`), path ownership, exact dependency SHAs, and an approved
228
- current fingerprint before it runs the single fast gate. `task-integrate` repeats
229
- the dependency and review checks. If integration stops on a conflict, resolve it
230
- inside the task worktree, run `task-resume`, obtain a fresh review, and then retry
231
- integration. A target rebase likewise requires a fresh review.
232
-
233
- The task commit runs project commit hooks once. The internal backup task-branch
234
- push skips the pre-push hook; the final target-branch push runs it once.
235
- `task-integrate` creates no second evidence commit.
236
-
237
- Successful integration removes the clean worktree and merged local branch, and
238
- deletes the merged remote task branch by default. Use `task-finish <TASK_ID>` to
239
- retry cleanup for one integrated task and `task-prune` for registered merged
240
- tasks left behind. Cleanup must refuse dirty worktrees and unmerged branches.
241
- Before forced task-worktree removal, recursively require initialized submodules
242
- to be clean and contain no unknown ignored data, local-only history, or
243
- unmirrored local branches. A submodule with another linked worktree or an
244
- unrefreshable remote blocks cleanup. Before forced worktree removal, persist its
245
- refs and reflog commits in the managed Git-common-dir submodule recovery store.
246
- Do not run submodule deinit from a linked task worktree because its shared config
247
- can alter the primary checkout.
248
- `task-start` enables worktree-scoped Git config and seeds task-local submodule
249
- URLs. After changing `.gitmodules`, run `task-submodule-sync` before submodule
250
- initialization or sync. Resolve relative URLs with Git's current-worktree
251
- default-remote semantics. If shared/common submodule config changed after task
252
- start, report and stop without restoring or overwriting it. Residual module Git
253
- directories from a manual deinit or removal block forced cleanup until they are
254
- reinitialized or recovered.
255
- Integration must not update the primary checkout; pull it explicitly only when
256
- its user-owned state is ready.
257
-
258
- Before running `autocoding sync`, finish, integrate, and clean every registered
259
- v3.0 task with the currently installed 3.0.0 runtime. Sync atomically rejects the
260
- whole project batch while any schema-v1 task remains, including
261
- `--components skill`; it never guesses or auto-claims owned paths during upgrade.
262
-
263
- ## Subagent orchestration
264
-
265
- The main agent owns the task graph and Git lifecycle. Before dispatch, decompose
266
- the request into bounded units and record for each unit its task ID, role, scope,
267
- dependencies, acceptance criteria, and expected result. Run only independent
268
- work concurrently.
269
-
270
- 1. Run justified `explorer`, `docs_researcher`, and `browser_debugger` roles in
271
- parallel for read-only discovery. Browser work collects reproduction evidence;
272
- it is not post-push acceptance.
273
- 2. The main agent merges discovery evidence and decides the design, dependency
274
- graph, and non-overlapping path ownership.
275
- 3. Assign each independent write unit to exactly one `fixer`. A single serial
276
- fixer may use a clean current branch; dirty checkouts and parallel fixers each
277
- require a registered task branch/worktree. Never run two writers in one worktree.
278
- Do not let the main agent edit a fixer-owned worktree concurrently.
279
- 4. Parallelize fixer units only when they are dependency-free and their owned
280
- paths do not overlap. Complete implementation → review → gate/integration for
281
- the current dependency wave, then start the next wave only after its
282
- prerequisites are integrated into the target branch.
283
- 5. A read-only `reviewer` checks a stable diff after implementation and binds its
284
- verdict to the diff fingerprint. Blocking findings return to the owning fixer;
285
- any edit invalidates approval and requires re-review. Record the final verdict
286
- with `task-review` before handing the writer lease back to the main agent.
287
- 6. Subagents never commit, push, integrate, clean branches, or run the project
288
- gate. Their versioned result contains node/task/role, base SHA, status,
289
- dependencies, owned and changed paths, diff fingerprint, evidence, findings,
290
- verdict, risks, and next owner. Read-only roles return no changed paths.
291
- 7. After all required results return, use `task-handoff` to transfer the lease to
292
- the main agent. The main agent alone runs `commit-push` for each write task,
293
- which validates the current review and runs the single fast gate, then
294
- integrates in dependency order, pushes, and cleans all temporary branches and
295
- worktrees.
296
-
297
- `classify` emits a machine-readable `agent_plan` with this stage and ownership
298
- contract. Micro tasks remain main-agent-only unless useful independent work
299
- clearly exceeds delegation overhead. If the client cannot run subagents, execute
300
- the same stages sequentially and check for a clean current branch before creating
301
- any temporary worktree.
302
-
303
- ## Default workflow
304
-
305
- Development:
306
-
307
- 1. Read this file and the active task entry.
308
- 2. Analyze and decompose the request; run or reason through `classify --scope auto`.
309
- 3. Fan out independent discovery roles, then fan their results back into one main-agent design.
310
- 4. Record the smallest useful design note; create DD/ADR only when indicated.
311
- 5. Use the clean current branch for one serial writer; only dirty or parallel work starts registered worktrees with explicit owned paths, dependency SHAs, and writer leases.
312
- 6. Review stable diffs, record their fingerprints with `task-review`, and return findings to their owning fixers until approved.
313
- 7. Hand the writer lease to the main agent. The main agent runs `commit-push` for each write task; it enforces ownership,
314
- dependency, lease, and current-review gates, executes the one changed-scope fast gate, records
315
- `DEV-CLOSED`, commits, and pushes the task branch.
316
- 8. The main agent runs `task-integrate` in dependency order to push the target branch and clean the task worktree and
317
- merged task branches. This completes the push stage and ends the coding task.
318
-
319
- Do not wait for, poll, diagnose, or record Jenkins/build/deploy results after the
320
- push. The project owner performs real acceptance separately.
321
-
322
- The authoritative order is analysis, decomposition, design, development, fast
323
- local gate, then push. Do not mark unexecuted external checks as PASS.
324
-
325
- ## Structure policy
326
-
327
- Generic structure checks are advisory by default. They may report file size,
328
- function size, or regex-based import-boundary signals, but they block only when
329
- `structure.enforcement: blocking` is explicitly configured. Repository-native
330
- architecture and compiler/test evidence take precedence over generic naming
331
- heuristics. Reuse existing helpers and avoid adding unrelated responsibilities
332
- to already large files; do not split code merely to satisfy a line count.
333
-
334
- Use `baseline init --write --update-config` for repository-wide structural work.
335
- Accepted historical debt belongs in the generated baseline/backlog.
336
-
337
- ## Core and optional documents
338
-
339
- The default scaffold contains only this file, taskbook, and closure log. Create
340
- specialized documents when the task needs them:
127
+ - A clean checkout with one writer uses the current branch directly.
128
+ - A checkout that already contains unrelated changes uses an isolated task
129
+ branch/worktree.
130
+ - Two or more concurrent writers always use separate worktrees with
131
+ non-overlapping owned paths.
132
+ - A task that produces no diff creates no commit or push; direct mode creates no
133
+ temporary branch at all.
134
+ - Never restore, reset, stash, clean, or otherwise modify unknown user or task
135
+ changes.
136
+
137
+ `task-start` implements this decision. Use `--isolated` to force a worktree and
138
+ `--review-required` when a project rule or the task risk requires fingerprinted
139
+ review:
341
140
 
342
141
  ```bash
343
- python3 docs/tools/autopipeline/ap.py scaffold api --write
344
- python3 docs/tools/autopipeline/ap.py scaffold design --write
345
- python3 docs/tools/autopipeline/ap.py scaffold architecture --write
346
- python3 docs/tools/autopipeline/ap.py scaffold review --write
347
- python3 docs/tools/autopipeline/ap.py scaffold testing --write
348
- python3 docs/tools/autopipeline/ap.py scaffold deployment --write
349
- python3 docs/tools/autopipeline/ap.py scaffold bugs --write
350
- python3 docs/tools/autopipeline/ap.py scaffold all --write
142
+ python3 docs/tools/autopipeline/ap.py task-start T0001 \
143
+ --owned-path src --writers 1
144
+ python3 docs/tools/autopipeline/ap.py commit-push T0001 --msg "T0001: summary"
351
145
  ```
352
146
 
353
- `baseline init` and `gen-summary` generate their outputs directly. The active
354
- taskbook, closure log, and top-level DD files are bounded working ledgers; use
355
- `docs-ledger-archive --plan` before archiving closed history.
147
+ In direct mode, `commit-push` runs the final gate, commits, and pushes the current
148
+ target branch. In isolated mode it pushes the task branch; the main agent then
149
+ runs `task-integrate`, which pushes the target branch and removes clean merged
150
+ temporary branches/worktrees. `task-prune` removes only safely merged leftovers.
356
151
 
357
- ## Common commands
152
+ ## Real changed-scope validation
358
153
 
359
- ```bash
360
- python3 docs/tools/autopipeline/ap.py doctor
361
- python3 docs/tools/autopipeline/ap.py classify --scope auto
362
- python3 docs/tools/autopipeline/ap.py light-gate --scope auto --explain
363
- python3 docs/tools/autopipeline/ap.py structure-check --scope auto
364
- python3 docs/tools/autopipeline/ap.py task-start <TASK_ID> --owned-path <PATH>
365
- python3 docs/tools/autopipeline/ap.py task-status <TASK_ID>
366
- python3 docs/tools/autopipeline/ap.py task-submodule-sync <TASK_ID>
367
- python3 docs/tools/autopipeline/ap.py task-review <TASK_ID> --verdict approved --diff-fingerprint <SHA256>
368
- python3 docs/tools/autopipeline/ap.py task-handoff <TASK_ID> --from <WRITER> --to <WRITER>
369
- python3 docs/tools/autopipeline/ap.py commit-push <TASK_ID> --msg "<TASK_ID>: <summary>"
370
- python3 docs/tools/autopipeline/ap.py task-integrate <TASK_ID>
371
- python3 docs/tools/autopipeline/ap.py task-resume <TASK_ID>
372
- python3 docs/tools/autopipeline/ap.py task-finish <TASK_ID>
373
- python3 docs/tools/autopipeline/ap.py task-prune
374
- python3 docs/tools/autopipeline/ap.py docs-ledger-check
375
- python3 docs/tools/autopipeline/ap.py gate-profile
154
+ `risk.rules` classifies work. `validation.routes` executes checks. Do not mix
155
+ these responsibilities.
156
+
157
+ ```yaml
158
+ commands:
159
+ backend_fast: "cd backend && go test ./internal/orders/..."
160
+ frontend_fast: "cd frontend && npm run test:changed"
161
+
162
+ validation:
163
+ on_unmapped: error
164
+ routes:
165
+ - name: backend
166
+ paths: ["backend/**", "contracts/**"]
167
+ commands: [backend_fast]
168
+ - name: frontend
169
+ paths: ["frontend/**"]
170
+ commands: [frontend_fast]
376
171
  ```
377
172
 
378
- `light-gate` is available for an explicit gate-diagnostic task. Do not add it to
379
- normal closure. Run exactly one configured changed gate directly on a clean
380
- serial branch or through `commit-push` for isolated work.
173
+ Every changed path may match multiple routes. The runtime collects all commands
174
+ in configuration order, removes duplicates, and executes each once. Unmapped
175
+ code fails before staging. Documentation-only changes may use the built-in diff
176
+ check. `git diff --check` is always an additional hygiene check and never counts
177
+ as business-code validation.
178
+
179
+ Inspect coverage without running the commands:
180
+
181
+ ```bash
182
+ python3 docs/tools/autopipeline/ap.py validation-map-check --path backend/foo.go
183
+ python3 docs/tools/autopipeline/ap.py validation-map-check --tracked
184
+ ```
381
185
 
382
- The launcher under `docs/tools/autopipeline` delegates to the single runtime in
383
- `.agents/skills/auto-coding-skill/scripts`; do not duplicate that runtime.
186
+ Focused tests may be run and rerun during implementation. The "one gate" rule
187
+ means one final routed closure gate after the diff is stable; it does not prohibit
188
+ reasonable test/fix/retest cycles. Do not install dependencies automatically.
189
+ If the final routed command needs a dependency already locked by the repository,
190
+ restore the locked dependency once and retry that same command only.
191
+
192
+ ## Design, review, and subagents
193
+
194
+ - Create DD/ADR only for lasting cross-module, API, data, security, deployment,
195
+ or key user-flow decisions.
196
+ - Ordinary work defaults to the main agent. Delegate only independent discovery
197
+ or implementation units whose latency or expertise benefit exceeds dispatch
198
+ cost.
199
+ - Read-only explorer/docs/browser roles may run in parallel when their questions
200
+ are independent.
201
+ - Parallel fixers require separate worktrees and non-overlapping owned paths.
202
+ - Require reviewer fingerprint approval for high-risk, cross-module, explicitly
203
+ configured, or parallel integration work; ordinary serial fixes do not require it.
204
+ - The main agent alone owns the final gate, commit, push, ordered integration,
205
+ and temporary-branch cleanup.
206
+
207
+ ## Lightweight records
208
+
209
+ Ordinary tasks do not create taskbook, closure Markdown, evidence JSONL, or active
210
+ task documents. Isolated-task coordination state lives under the Git common
211
+ directory and gate timing lives under `.local/auto-coding-skill`. Existing project
212
+ ledgers are user data: upgrades preserve them but do not require or update them.
213
+ Use `record-closure`, documentation scaffolds, or a baseline only when the user or
214
+ task genuinely needs a durable artifact.
215
+
216
+ Historical debt does not block product work. Block only new or worsened P0/P1
217
+ issues; keep accepted debt in an optional backlog and address it through explicit
218
+ governance tasks.
219
+
220
+ ## Access and explicit diagnostics
221
+
222
+ Fill all URL, username, and password fields under `access.*` during project
223
+ initialization. Direct plaintext values are allowed. The generic workflow neither
224
+ forbids them nor rewrites a project's credential policy.
225
+
226
+ Full regression, structure scans, Docker, Jenkins, API checks, target verification,
227
+ and deployment remain explicit diagnostic commands. Do not run them during normal
228
+ closure unless the user explicitly requests that diagnostic scope.
384
229
 
385
230
  <!-- auto-coding-skill:managed-workflow:end -->
@@ -1,16 +1,29 @@
1
- <!-- auto-coding-skill:managed-agents:start version=3.0.3 -->
1
+ <!-- auto-coding-skill:managed-agents:start version=4.0.0 -->
2
2
 
3
- Follow `docs/ENGINEERING.md` strictly. Source of truth: `docs/ENGINEERING.md`.
4
- At task start, use `ap.py classify --scope auto` when changed-file impact is not obvious.
5
- Before code changes, read the configured structure standard and place new code in the correct layer; do not add new responsibilities to already-large files.
6
- Obey `structure.layer_rules` import boundaries unless a DD/ADR explicitly records the exception.
7
- Before optimization reviews, read the project health baseline and optimization backlog, and report only new, worsened, unrecorded P0/P1, or upgraded-priority items.
8
- Route work through already available MCP servers, installed skills, plugins, and app connectors when they provide current docs, design context, browser evidence, GitHub/CI state, security review, or artifact rendering.
9
- Use the `.agents/agents` role model when available. The main agent decomposes and fans out independent explorer/docs_researcher/browser_debugger work, then fans evidence into one design. For one serial writer, use the current branch when its checkout is clean; create registered task branches/worktrees only when that checkout is already dirty or multiple writers run concurrently. Give every writer explicit non-overlapping paths, route stable diffs through reviewer, then let the main agent run the one fast gate, commit, integrate, push, and clean temporary branches. Never run two writers in one worktree; start dependent write work only after its prerequisite is integrated.
10
- Use exactly one fast changed-scope local gate before pushing. Keep advisory structure, docs-ledger, full regression, Docker, API, Jenkins, and target checks outside the default development gate.
11
- Keep active docs ledgers small: `taskbook.md`, `closure-log.md`, and top-level `docs/design/T*.md` must be physically archived when over budget. Use `docs-ledger-archive --plan` before `--write`; `archive-index.md` is only navigation.
12
- Keep `docs/tasks/evidence.jsonl` and closure Markdown aligned with actual executed checks.
13
- Treat target-branch push as coding completion. Do not wait for, poll, diagnose, or fix Jenkins/build/deploy results unless the user starts a separate diagnostic task; the project owner performs acceptance.
14
- Use the direct URL, username, and password values configured under `access.*` in `docs/ENGINEERING.md`; do not invent credentials.
3
+ Follow `docs/ENGINEERING.md` for project facts, risk rules, and validation routes.
4
+ Use the delivery-first flow: analyze, decompose, design only when needed, develop,
5
+ run one final changed-scope gate, then commit and push. Push ends coding; do not
6
+ wait for Jenkins, deployment, or owner acceptance unless the user opens a
7
+ diagnostic task.
8
+
9
+ Before writing, inspect the checkout. A clean single-writer task stays on the
10
+ current branch. Pre-existing unrelated changes or multiple writers require one
11
+ isolated worktree per writer with non-overlapping owned paths. If no diff is
12
+ produced, do not commit or push and do not create a temporary branch. Never
13
+ restore, reset, stash, clean, or overwrite unknown changes.
14
+
15
+ Delegate only when independent work has a clear benefit. Ordinary tasks default
16
+ to the main agent. Require fingerprinted review for high-risk, cross-module,
17
+ parallel, or explicitly configured work. The main agent owns the final gate,
18
+ commit, push, ordered integration, and cleanup.
19
+
20
+ Run all matching `validation.routes` commands once in stable order. Unmapped code
21
+ must fail; docs-only changes may use the built-in diff check. Focused tests may be
22
+ rerun during development, but do not repeat expensive full gates. Do not create
23
+ taskbook, closure, evidence, or design documents unless the task genuinely needs
24
+ a durable artifact.
25
+
26
+ Use configured `access.*` values when needed. Plaintext credentials are allowed;
27
+ do not invent or echo them unnecessarily.
15
28
 
16
29
  <!-- auto-coding-skill:managed-agents:end -->