@elvis1513/auto-coding-skill 4.3.6 → 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.
- package/README.md +26 -748
- package/cli/assets/managed-install.json +38 -1
- package/cli/assets/skill/SKILL.md +15 -98
- package/cli/assets/skill/data/templates/ENVIRONMENT.md +30 -0
- package/cli/assets/skill/data/templates/architecture/.gitkeep +1 -0
- package/cli/assets/skill/data/templates/bridges/AGENTS.md +39 -143
- package/cli/assets/skill/data/templates/deployment/.gitkeep +1 -0
- package/cli/assets/skill/data/templates/design/.gitkeep +1 -0
- package/cli/assets/skill/data/templates/interfaces/.gitkeep +1 -0
- package/cli/assets/skill/data/templates/product/.gitkeep +1 -0
- package/cli/src/index.js +127 -2589
- package/cli/src/sync-assets.js +38 -95
- package/package.json +9 -17
- package/cli/assets/agents/browser-debugger.toml +0 -27
- package/cli/assets/agents/docs-researcher.toml +0 -18
- package/cli/assets/agents/explorer.toml +0 -26
- package/cli/assets/agents/fixer.toml +0 -29
- package/cli/assets/agents/reviewer.toml +0 -50
- package/cli/assets/skill/data/contracts/orchestration-v1.schema.json +0 -567
- package/cli/assets/skill/data/policies/feedback-resolutions-v1.json +0 -68
- package/cli/assets/skill/data/policies/workflow-migrations-v1.json +0 -657
- package/cli/assets/skill/data/templates/ENGINEERING.md +0 -115
- package/cli/assets/skill/data/templates/tools/ap.py +0 -16
- package/cli/assets/skill/requirements.txt +0 -1
- package/cli/assets/skill/scripts/ap.py +0 -17630
- package/cli/assets/skill/scripts/core.py +0 -510
- package/cli/assets/skill/scripts/http_checks.py +0 -92
- package/cli/assets/skill/scripts/install_integrity.py +0 -427
- package/cli/assets/skill/scripts/scaffold_templates.py +0 -325
- package/cli/tests/ap_concurrency_test.py +0 -3520
- package/cli/tests/ap_profiles_test.py +0 -3261
- package/cli/tests/core_shell_test.py +0 -345
- package/cli/tests/index.test.js +0 -2967
|
@@ -1 +1,38 @@
|
|
|
1
|
-
{
|
|
1
|
+
{
|
|
2
|
+
"schema_version": 2,
|
|
3
|
+
"skill_version": "5.0.0",
|
|
4
|
+
"asset_files": [
|
|
5
|
+
{
|
|
6
|
+
"path": "SKILL.md",
|
|
7
|
+
"sha256": "2f254d4e45ca5d2965cb1ba3fbeffe420f98f7f6da4451bde4e5ba94c172cf81"
|
|
8
|
+
},
|
|
9
|
+
{
|
|
10
|
+
"path": "data/templates/ENVIRONMENT.md",
|
|
11
|
+
"sha256": "d049902afb929e3ec7883f177762dbaff7dc2cbe526b3e9b8163bff86ba03fcd"
|
|
12
|
+
},
|
|
13
|
+
{
|
|
14
|
+
"path": "data/templates/architecture/.gitkeep",
|
|
15
|
+
"sha256": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b"
|
|
16
|
+
},
|
|
17
|
+
{
|
|
18
|
+
"path": "data/templates/bridges/AGENTS.md",
|
|
19
|
+
"sha256": "af9058ff27707506be1f12a8b634617d906f62462cc1069cd6c853c069aef9af"
|
|
20
|
+
},
|
|
21
|
+
{
|
|
22
|
+
"path": "data/templates/deployment/.gitkeep",
|
|
23
|
+
"sha256": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b"
|
|
24
|
+
},
|
|
25
|
+
{
|
|
26
|
+
"path": "data/templates/design/.gitkeep",
|
|
27
|
+
"sha256": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b"
|
|
28
|
+
},
|
|
29
|
+
{
|
|
30
|
+
"path": "data/templates/interfaces/.gitkeep",
|
|
31
|
+
"sha256": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b"
|
|
32
|
+
},
|
|
33
|
+
{
|
|
34
|
+
"path": "data/templates/product/.gitkeep",
|
|
35
|
+
"sha256": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b"
|
|
36
|
+
}
|
|
37
|
+
]
|
|
38
|
+
}
|
|
@@ -1,108 +1,25 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: auto-coding-skill
|
|
3
|
-
description:
|
|
3
|
+
description: General project guidance and a lightweight documentation layout. It provides context files and optional places to record durable project knowledge, without imposing a development workflow.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
6
|
# Auto Coding Skill
|
|
7
7
|
|
|
8
|
-
|
|
8
|
+
Read `AGENTS.md`, then use the project documentation only when it helps the
|
|
9
|
+
current request:
|
|
9
10
|
|
|
10
|
-
|
|
11
|
-
|
|
12
|
-
`docs/
|
|
13
|
-
|
|
11
|
+
- `docs/ENVIRONMENT.md` for local/runtime facts and project-approved access data.
|
|
12
|
+
- `docs/product/` for durable product context and decisions.
|
|
13
|
+
- `docs/architecture/`, `docs/design/`, `docs/interfaces/`, and
|
|
14
|
+
`docs/deployment/` for existing topic-specific knowledge.
|
|
14
15
|
|
|
15
|
-
|
|
16
|
+
This Skill has no required task protocol. It does not require classification,
|
|
17
|
+
Gates, Reviews, worktrees, designs, tests, commits, or deployment checks.
|
|
16
18
|
|
|
17
|
-
|
|
19
|
+
Write documentation only when the user asks for it or when a durable fact,
|
|
20
|
+
contract, decision, test strategy, or operating procedure would otherwise be
|
|
21
|
+
lost. Choose the relevant existing folder; do not create routine task records,
|
|
22
|
+
review records, or duplicate summaries.
|
|
18
23
|
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
Read-only questions need no workflow command. Clean serial edits and terminal
|
|
22
|
-
maintenance may follow `AGENTS.md` directly. When impact or concurrency is unclear, run:
|
|
23
|
-
|
|
24
|
-
```bash
|
|
25
|
-
python3 docs/tools/autopipeline/ap.py classify --scope auto \
|
|
26
|
-
--planned-path <PATH> --intent "<intent>" [--writers <N>] \
|
|
27
|
-
[--task-kind read_only|change|terminal_maintenance] [--claim-direct]
|
|
28
|
-
```
|
|
29
|
-
|
|
30
|
-
Run `mechanism_plan.required`; select `optional_when_beneficial` only when its
|
|
31
|
-
value exceeds coordination cost, and keep `forbidden` off absent user override.
|
|
32
|
-
`classify` is a fail-closed snapshot: inspect `repo`, `workspace_dirty`,
|
|
33
|
-
`dirty_paths`, and `active_writer`, and reclassify if state changes before writing.
|
|
34
|
-
Same-owner claims conflict unless continuing the exact ID; `change_nature` never bypasses a rule or review.
|
|
35
|
-
|
|
36
|
-
Use the registered lifecycle only when classification requires isolation/fingerprinted review or the user explicitly requests it:
|
|
37
|
-
|
|
38
|
-
```bash
|
|
39
|
-
python3 docs/tools/autopipeline/ap.py task-start T0001 \
|
|
40
|
-
--owned-path src [--isolated] [--review-required] [--continue-direct --direct-claim <ID>]
|
|
41
|
-
# Before the first commit, expand an active task instead of restarting it:
|
|
42
|
-
python3 docs/tools/autopipeline/ap.py task-scope-add T0001 --owned-path config/new-scope
|
|
43
|
-
# Run a read-only Reviewer under its fixed deadline and record the bound result:
|
|
44
|
-
python3 docs/tools/autopipeline/ap.py review-run T0001 --reviewer <REVIEWER_ID> --json
|
|
45
|
-
python3 docs/tools/autopipeline/ap.py commit-push T0001 --msg "T0001: summary"
|
|
46
|
-
# Isolated tasks only:
|
|
47
|
-
python3 docs/tools/autopipeline/ap.py task-integrate T0001
|
|
48
|
-
```
|
|
49
|
-
|
|
50
|
-
Each parallel writer uses a task ID, worktree, lease, dependency SHAs, and distinct
|
|
51
|
-
`owned_paths`. Main integrates, gates, pushes, and cleans. Validate delegated JSON
|
|
52
|
-
with `agent-contract-check`; parallel fixers require reclassification with
|
|
53
|
-
`--writers >1`.
|
|
54
|
-
|
|
55
|
-
Focused review uses `high`/150 seconds; deep review uses `xhigh`/360 seconds. `review-run` launches a separate read-only Codex process, streams allowlisted event metadata to a private Git-local log, stops the process group at deadline, and binds the result to the exact HEAD/scope/fingerprint.
|
|
56
|
-
No semantic event within 30 seconds retries the same assignment once; started analysis is never retried. `review-assignment` freezes every working-tree change into a mode-0600, assignment-bound, SHA-256 patch.
|
|
57
|
-
The Reviewer reads it with `review-artifact --file <assignment.json>`, never a reconstructed live diff. `review-assignment` alone needs another deadline-capable host; `agent-result-template` supplies all 16 fields.
|
|
58
|
-
|
|
59
|
-
After exhausted startup attempts or an analysis deadline, explicit user authorization may record a one-fingerprint `review-runtime-override` with identity, reference, reason, evidence, and `--confirm-runtime-bypass`.
|
|
60
|
-
It records `runtime-bypassed`, never `approved`, and cannot replace a substantive `blocked` or `changes-requested` result.
|
|
61
|
-
An active task created by 4.2.8-4.3.2 may instead use `review-runtime-retry` only
|
|
62
|
-
when its sole completed blocked result is the fixed Git-local `review-artifact`
|
|
63
|
-
permission defect with no findings. The lifecycle owner must supply the unchanged
|
|
64
|
-
fingerprint and fixed reason code from a strictly newer managed runtime; the
|
|
65
|
-
original Reviewer identity then runs once against a create-only audit and fresh
|
|
66
|
-
fixed deadline. This migration never renews an ordinary timeout or substantive
|
|
67
|
-
review result. A 4.3.5+ runtime may repair only an untouched `retry-v4.3.4` or its
|
|
68
|
-
exact non-substantive artifact-access block, preserving both audit generations.
|
|
69
|
-
|
|
70
|
-
## Close with the bounded routed gate
|
|
71
|
-
|
|
72
|
-
For unregistered clean serial work, run the final gate and then use normal Git:
|
|
73
|
-
|
|
74
|
-
```bash
|
|
75
|
-
python3 docs/tools/autopipeline/ap.py validation-map-check --path <PATH>
|
|
76
|
-
python3 docs/tools/autopipeline/ap.py light-gate --scope changed --explain
|
|
77
|
-
```
|
|
78
|
-
|
|
79
|
-
For a registered task, call only `commit-push`: it owns the final gate and reuses
|
|
80
|
-
an exact Git-local PASS only while content, base, scope, routes, commands, and
|
|
81
|
-
lease match. Matching manual passes are reusable; changed state is not.
|
|
82
|
-
|
|
83
|
-
Every changed code/config path must map to a real project command. The runtime
|
|
84
|
-
de-duplicates commands and enforces 120-second command and 180-second total
|
|
85
|
-
budgets; measured routes may override them. Timeouts narrow, never expand, scope.
|
|
86
|
-
Focused test/fix/retest remains allowed; commands inherit PATH in non-login Bash.
|
|
87
|
-
Size warnings block only with `structure.enforcement: blocking` and `structure.block_warnings: true` (default `false`).
|
|
88
|
-
|
|
89
|
-
Run full regression, Docker, Jenkins, deployment, browser/device writes, or target
|
|
90
|
-
acceptance only as an explicit diagnostic. A requested just-pushed failure
|
|
91
|
-
diagnosis continues in the same task without inventing another lifecycle.
|
|
92
|
-
|
|
93
|
-
## Keep state and upgrades safe
|
|
94
|
-
|
|
95
|
-
Machine task state, leases, fingerprints, and gate evidence stay Git-local.
|
|
96
|
-
Ordinary work creates no taskbook, closure, evidence JSONL, active-task, or
|
|
97
|
-
design artifact. Terminal ledger reconciliation validates and commits once.
|
|
98
|
-
|
|
99
|
-
```bash
|
|
100
|
-
autocoding init
|
|
101
|
-
python3 docs/tools/autopipeline/ap.py doctor
|
|
102
|
-
```
|
|
103
|
-
|
|
104
|
-
Finish registered tasks before changing versions. `autocoding init` idempotently
|
|
105
|
-
replaces managed defaults and constraints while preserving the project overlay
|
|
106
|
-
byte-for-byte after its one-time semantic migration. It installs canonical docs,
|
|
107
|
-
archives obsolete content outside active `docs/`, verifies the managed-install
|
|
108
|
-
manifest, and lets `doctor` validate the effective configuration.
|
|
24
|
+
`autocoding init` creates missing guidance, the environment file, and empty
|
|
25
|
+
topic directories. It never overwrites project documentation.
|
|
@@ -0,0 +1,30 @@
|
|
|
1
|
+
# Project Environment
|
|
2
|
+
|
|
3
|
+
> Project-owned operational context. Populate only information useful to this
|
|
4
|
+
> project. Plaintext usernames and passwords are permitted here when the project
|
|
5
|
+
> owner intentionally wants local development and diagnosis to use them. Do not
|
|
6
|
+
> invent, echo, or copy credentials outside the project.
|
|
7
|
+
|
|
8
|
+
## Local development
|
|
9
|
+
|
|
10
|
+
| Item | Value / command | Notes |
|
|
11
|
+
| --- | --- | --- |
|
|
12
|
+
| Repository root | | |
|
|
13
|
+
| Runtime/toolchain | | |
|
|
14
|
+
| Main services | | |
|
|
15
|
+
| Start command | | |
|
|
16
|
+
|
|
17
|
+
## Environments and access
|
|
18
|
+
|
|
19
|
+
| Environment | System / URL | Username | Password | Notes |
|
|
20
|
+
| --- | --- | --- | --- | --- |
|
|
21
|
+
| Development | | | | |
|
|
22
|
+
| Staging | | | | |
|
|
23
|
+
| Production | | | | |
|
|
24
|
+
|
|
25
|
+
## Operational facts
|
|
26
|
+
|
|
27
|
+
- Important configuration locations:
|
|
28
|
+
- Logs/metrics locations:
|
|
29
|
+
- External dependencies:
|
|
30
|
+
- Known constraints:
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
|
|
@@ -1,157 +1,53 @@
|
|
|
1
|
-
<!-- auto-coding-skill:managed-agents:start version=
|
|
2
|
-
#
|
|
1
|
+
<!-- auto-coding-skill:managed-agents:start version=5.0.0 -->
|
|
2
|
+
# General Engineering Guidance
|
|
3
3
|
|
|
4
|
-
|
|
5
|
-
`docs/project/` and configuration in `docs/project/auto-coding-skill.yaml`.
|
|
4
|
+
## Authority and scope
|
|
6
5
|
|
|
7
|
-
|
|
6
|
+
Use the current user request as the primary authority. Treat the current source,
|
|
7
|
+
tests, schemas, runtime configuration, and deployment configuration as the truth
|
|
8
|
+
for existing behavior. Documentation describes durable intent and operating
|
|
9
|
+
context; it must not override working code without an explicit decision.
|
|
8
10
|
|
|
9
|
-
|
|
11
|
+
Work only within the requested scope. Do not silently change deployment systems,
|
|
12
|
+
Jenkins, production data, credentials, infrastructure, or another developer's
|
|
13
|
+
work. When a shared checkout contains unrelated changes, preserve them and avoid
|
|
14
|
+
reset, clean, overwrite, or bulk formatting operations.
|
|
10
15
|
|
|
11
|
-
|
|
12
|
-
2. Code, tests, schemas, migrations, and runtime configuration for behavior that
|
|
13
|
-
exists now.
|
|
14
|
-
3. The effective configuration formed from managed `docs/ENGINEERING.md` defaults
|
|
15
|
-
plus the project overlay, then `docs/project/` for durable project facts.
|
|
16
|
-
4. Current interface documentation for intended contracts.
|
|
17
|
-
5. Accepted DD/ADR for lasting decisions.
|
|
18
|
-
6. Taskbooks, closures, reviews, and deployment records as history only.
|
|
16
|
+
## Development approach
|
|
19
17
|
|
|
20
|
-
|
|
21
|
-
|
|
18
|
+
Choose the smallest useful approach for the request. Work directly for ordinary
|
|
19
|
+
changes. Use a branch, worktree, tests, review, design note, or diagnostic only
|
|
20
|
+
when the task or repository makes it useful; none is required by this guidance.
|
|
22
21
|
|
|
23
|
-
|
|
22
|
+
Do not treat a test, build, lint, review, commit, push, Jenkins run, deployment,
|
|
23
|
+
or target check as mandatory completion ceremony. Run a check when the user asks,
|
|
24
|
+
when the project explicitly needs it, or when it is the proportionate way to
|
|
25
|
+
answer a technical question. State clearly what was and was not verified.
|
|
24
26
|
|
|
25
|
-
|
|
26
|
-
The default budget is:
|
|
27
|
+
## Documentation
|
|
27
28
|
|
|
28
|
-
|
|
29
|
-
| --- | --- |
|
|
30
|
-
| Read-only question | analysis only; no workflow command |
|
|
31
|
-
| Obvious clean serial edit | analysis, affected gate, commit/push |
|
|
32
|
-
| Terminal ledger/archive maintenance | targeted consistency check, one commit |
|
|
33
|
-
| Dirty checkout or concurrent writer | registered lifecycle and isolated worktree |
|
|
34
|
-
| High-risk or contract-crossing change | necessary design and independent review |
|
|
35
|
-
| Multiple independent writers | one task/worktree/lease per writer, ordered integration |
|
|
29
|
+
Read relevant project documentation before relying on project context:
|
|
36
30
|
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
|
|
42
|
-
|
|
31
|
+
- `docs/ENVIRONMENT.md`: local/runtime services, commands, endpoints, and access data.
|
|
32
|
+
- `docs/product/`: durable product context, boundaries, and decisions.
|
|
33
|
+
- `docs/architecture/`: architecture and ADRs.
|
|
34
|
+
- `docs/design/`: durable design notes that help future implementation.
|
|
35
|
+
- `docs/interfaces/`: public API, event, protocol, or integration contracts.
|
|
36
|
+
- `docs/deployment/`: runbooks and durable operating procedures.
|
|
43
37
|
|
|
44
|
-
|
|
45
|
-
|
|
38
|
+
Documentation is optional. Do not create a record for a routine edit, a transient
|
|
39
|
+
investigation, an ordinary review, or every commit. Record something only when the
|
|
40
|
+
user asks or when the information will remain useful after the immediate task.
|
|
41
|
+
Put it in the narrowest applicable folder and update an existing document instead
|
|
42
|
+
of creating parallel summaries.
|
|
46
43
|
|
|
47
|
-
|
|
48
|
-
owner acceptance automatically. A user-requested diagnosis of the just-pushed
|
|
49
|
-
failure may continue in the same conversation/task without a second lifecycle.
|
|
44
|
+
## Environment and credentials
|
|
50
45
|
|
|
51
|
-
|
|
52
|
-
|
|
53
|
-
|
|
54
|
-
|
|
55
|
-
|
|
56
|
-
|
|
57
|
-
- Every delegated fixer or parallel writer owns a task ID/worktree, writer lease,
|
|
58
|
-
non-overlapping `owned_paths`, and prerequisite commit SHAs.
|
|
59
|
-
- Before its first commit, an active task may add explicit paths with
|
|
60
|
-
`task-scope-add`; the command checks leases/conflicts, only raises risk, and
|
|
61
|
-
invalidates prior review and final-gate state. Never restart a no-diff task just
|
|
62
|
-
to enlarge its declared scope.
|
|
63
|
-
- Never let multiple writing agents share a checkout. Never restore, reset, stash,
|
|
64
|
-
clean, overwrite, or commit unknown changes.
|
|
65
|
-
- No diff means no commit, push, or temporary branch.
|
|
66
|
-
- Main integrates in dependency order and removes only clean, merged temporary state.
|
|
67
|
-
- Do not sync or upgrade while any registered task is active.
|
|
68
|
-
|
|
69
|
-
## Design, agents, and review
|
|
70
|
-
|
|
71
|
-
- Micro and standard tasks stay main-agent-only unless the model identifies an
|
|
72
|
-
independent question whose latency or expertise benefit exceeds coordination
|
|
73
|
-
cost.
|
|
74
|
-
- Create DD/ADR only for lasting cross-module, API, data, security, deployment,
|
|
75
|
-
or key user-flow decisions.
|
|
76
|
-
- Use read-only explorer/docs/browser agents only for independent questions.
|
|
77
|
-
- Use parallel fixers only for bounded, dependency-free, non-overlapping units.
|
|
78
|
-
- Reclassify before parallel fixers; single-writer plans grant no implicit writes.
|
|
79
|
-
- Validate delegated assignment/result JSON with `agent-contract-check` before use.
|
|
80
|
-
- Require fingerprinted review only for path/rule-confirmed high-risk,
|
|
81
|
-
contract-crossing, parallel, or configured work; intent words are candidates only.
|
|
82
|
-
- Use `high`/150 seconds for focused review and `xhigh`/360 seconds for deep parallel,
|
|
83
|
-
cross-module, or sensitive review. Retry once only if no semantic event arrives
|
|
84
|
-
within 30 seconds; never retry started analysis.
|
|
85
|
-
- Run `python3 docs/tools/autopipeline/ap.py review-run <TASK> --reviewer <ID> --json`.
|
|
86
|
-
It runs a separate read-only identity, streams allowlisted metadata privately,
|
|
87
|
-
stops the process group at deadline, and binds a mode-0600 SHA-256 patch to the
|
|
88
|
-
assigned HEAD, scope, identity, and fingerprint. Reviewers read the frozen patch
|
|
89
|
-
with `review-artifact`; `review-assignment` alone needs a deadline-capable host and cannot stop an in-app subagent.
|
|
90
|
-
- `agent-result-template` supplies all 16 fields and `review-run` normalizes presentation.
|
|
91
|
-
- Runtime exhaustion needs explicit `review-runtime-override` authorization; it records
|
|
92
|
-
`runtime-bypassed`, never approval, and cannot replace a substantive blocked result.
|
|
93
|
-
- `review-runtime-retry` is a narrow migration only for active 4.2.8-4.3.2 tasks
|
|
94
|
-
consumed solely by the fixed Git-local artifact permission defect. It requires
|
|
95
|
-
the lifecycle owner, unchanged fingerprint, original Reviewer, a strictly newer
|
|
96
|
-
managed runtime, and the fixed reason code; it preserves all old evidence and
|
|
97
|
-
permits one audited attempt. A 4.3.5+ runtime may repair only an untouched
|
|
98
|
-
`retry-v4.3.4` or its exact non-substantive artifact-access block, with both
|
|
99
|
-
audit generations retained. Never use it for timeouts, findings, changed scope,
|
|
100
|
-
user overrides, or general Reviewer failures.
|
|
101
|
-
- Block only new or worsened in-scope P0/P1 defects. Adjacent debt is a follow-up;
|
|
102
|
-
semantic changes invalidate approval, while mechanical docs may get a targeted recheck.
|
|
103
|
-
|
|
104
|
-
## Bounded real validation
|
|
105
|
-
|
|
106
|
-
- `risk.rules` controls reasoning/review depth. `validation.routes` alone selects
|
|
107
|
-
executable closure checks.
|
|
108
|
-
- Every changed code/config path must match one or more explicit routes. Run every
|
|
109
|
-
matched command once in stable order and de-duplicate references.
|
|
110
|
-
- Unmapped code or blank commands fail before staging. Documentation-only work may
|
|
111
|
-
use the built-in diff/format check. `git diff --check` is hygiene, not business
|
|
112
|
-
validation.
|
|
113
|
-
- Focused tests may run and rerun during development. Run one final routed gate
|
|
114
|
-
after the diff is stable.
|
|
115
|
-
- For registered tasks, `commit-push` owns that final gate; do not run a separate
|
|
116
|
-
`light-gate` first. A strict Git-local PASS receipt may be reused only while the
|
|
117
|
-
content/index, base, scope, route/command plan, and writer lease still match.
|
|
118
|
-
Unregistered clean serial work runs `light-gate` before normal Git closure.
|
|
119
|
-
- Final gate budgets are 120 seconds per command and 180 total; projects may raise
|
|
120
|
-
measured budgets or set smaller route timeouts. Timeouts narrow, never expand,
|
|
121
|
-
scope. Size warnings block only with `structure.enforcement: blocking` and
|
|
122
|
-
`structure.block_warnings: true` (default `false`).
|
|
123
|
-
- Do not proactively install dependencies. Only after the selected route fails
|
|
124
|
-
because a repository-locked dependency is absent may it be restored once and
|
|
125
|
-
only that route retried.
|
|
126
|
-
- Full regression, repository-wide scans, builds, Docker, Jenkins, deployment,
|
|
127
|
-
live-device writes, browser/API acceptance, and target checks are explicit
|
|
128
|
-
diagnostics, not normal closure.
|
|
129
|
-
- External-system writes require the project safety rule or user confirmation;
|
|
130
|
-
use fixtures/simulators when the project supplies them.
|
|
131
|
-
|
|
132
|
-
## Documentation and access
|
|
133
|
-
|
|
134
|
-
- This AGENTS file is the single behavioral protocol. The installed `SKILL.md`
|
|
135
|
-
contains invocation guidance. `docs/ENGINEERING.md` is managed default policy;
|
|
136
|
-
`docs/project/auto-coding-skill.yaml` is the higher-priority project overlay,
|
|
137
|
-
and the remaining `docs/project/` files own durable facts.
|
|
138
|
-
- `autocoding init` owns the active `docs/` directory topology and managed
|
|
139
|
-
templates. Put durable facts into the installed project files. Architecture,
|
|
140
|
-
ADR, interface, DD, review, and deploy-record artifacts may use their designated
|
|
141
|
-
directories; unrelated directories are archived outside active docs on upgrade.
|
|
142
|
-
- The managed-install manifest owns only its declared files and exact namespaces.
|
|
143
|
-
Never treat the whole `.agents` or `docs` tree as a release-package mirror.
|
|
144
|
-
- Ordinary work creates no taskbook, closure Markdown, evidence JSONL, active-task
|
|
145
|
-
document, or design file. Machine coordination/evidence stays in Git
|
|
146
|
-
common/local state and cannot change the reviewed diff.
|
|
147
|
-
- After diagnosing the project/Skill boundary, record a probable shared-Skill
|
|
148
|
-
defect or gap from `docs/skill-feedback/_TEMPLATE-SKILL-FEEDBACK.md`; keep
|
|
149
|
-
project-only configuration in `docs/project/auto-coding-skill.yaml`, redact
|
|
150
|
-
sensitive/raw evidence, and never mutate a frozen Reviewer diff merely to add
|
|
151
|
-
feedback. Periodically run `autocoding feedback --projects . --json`; after an
|
|
152
|
-
upgrade, verify and close/delete fixed reports or update current regressions.
|
|
153
|
-
- Pure ledger/archive reconciliation is terminal: validate, commit once, and stop.
|
|
154
|
-
- Fill configured project/Jenkins/GitLab/Nexus access fields during initialization.
|
|
155
|
-
Plaintext values are allowed; do not invent or unnecessarily echo credentials.
|
|
46
|
+
`docs/ENVIRONMENT.md` is project-owned operational context. The project owner may
|
|
47
|
+
store approved plaintext usernames and passwords there when that makes local
|
|
48
|
+
development and diagnosis practical. Never invent credentials, copy them outside
|
|
49
|
+
the project, place them in chat output, or expose them in logs. Keep endpoints,
|
|
50
|
+
commands, service names, and access instructions current when they materially
|
|
51
|
+
change.
|
|
156
52
|
|
|
157
53
|
<!-- auto-coding-skill:managed-agents:end -->
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
|