new-repo-ai-policy
暂无简介
- License 许可
- GPL-3.0
- 最近更新
- 2026-11-13
- Stars
- ★ 0
- Forks
- 0
03 SKILL.md 正文
new-repo-ai-policy
Owns the repo's AI-POLICY.md and its AI-POLICY-SHORT.md companion.
Editing them any other way — "just adding the model we used", writing
disclosure rules into a manuscript or a skill — is the RED-baseline failure
this skill exists to prevent (baselines-phase1.md, scenario D; design D3:
acceptance is a human act).
Trinity role: AI-POLICY.md (source of truth) + latex-ai-statement
(read-only auditor) + this skill (sole write path).
Procedure
Follow subdomains/repo-docs/AMENDMENT.md end to end: resolve the repo's
AI-POLICY.md (on miss, instantiate
subdomains/repo-docs/templates/AI-POLICY.md — the instantiation is itself
the proposal) → draft the structured proposal → human gate, mandatory →
apply to AI-POLICY.md (and its SHORT companion) only, with a Change Log
row naming the approver → verify (below) → conservative, pathspec-scoped
commit only with explicit authority.
This skill's specifics
- Declared-tools rows are the common case. Adding a tool, model, or
skill to the table is a one-row proposal, but it obliges a
.bibentry and a pinpoint citation wherever the tool is used (AI2, AI3) — name that in Downstream. - Placement changes (AI12) ripple into every manuscript in the repo; a journal requirement that overrides placement is recorded as such, not as a weakening of the rule.
- Pricing conventions live in AI16 and are implemented by
update-tokens; a proposal changing the rate source or the estimate-labelling convention must say what happens to existing rows (never silently re-priced). - Keep the SHORT companion in step. Every rule added, reworded, or
deprecated updates its indexed line in
AI-POLICY-SHORT.mdin the same amendment. The companion is an index, never a second source of truth.
Verification (Step 4)
First run the coverage check —
python3 subdomains/repo-docs/scripts/check-ai-coverage.py. It asserts that
every rule marked [C], [R], or [F] has an owner that names it: [C] a
row in latex-ai-statement's table, [R] a mention in update-ai-usage or
update-tokens, [F] a matching line in the SHORT index. Three review
rounds each shipped a rule whose enforcement point did not mention it; this
is the check that ends that. A non-zero exit blocks the amendment.
Then the manuscript pass. Run latex-ai-statement in
policy-verification mode on the amended AI-POLICY.md, then on each .tex
in the repo, and report the manuscript count — a repo with no manuscript
reports "checked 0 manuscripts", never a silent pass. AI15/AI16 amendments
touch the records, not a .tex: verify those against update-ai-usage and
update-tokens instead.
Floors
The template marks floors [F]. Read them there — this skill does not
restate the set, because a second copy is how the list drifts. A proposal
that weakens or deletes an [F] rule is refused, not presented; a repo whose
instantiated copy predates a floor is not exempt, and the remedy is an
amendment adding it.
Refuse and escalate
Route to the owning skill, never to the document — editing another domain's policy in place is the sole-write-path violation this trinity exists to prevent.
| Finding belongs to | Route to |
|---|---|
LX floor (subdomains/latex/POLICY.md) |
new-latex-policy |
Repo STYLE.md / LAYOUT.md / AGENTS.md |
new-repo-style-policy / new-repo-layout-policy / new-repo-agents-policy |
Repo ADR.md / LATEX.md |
new-repo-adr-policy / new-repo-latex-policy |
What a task used or cost (ai-usage.md, tokens.md) |
update-ai-usage / update-tokens — records, not contract |
| The pack templates themselves | agents decide (delegated by the repository owner, 2026-09-19), subject to the rule budget in the template header — no skill owns these, so propose durably, tighten the wording, and record the reasoning in the Change Log |
04 同仓库更多 Skills
writing-skills
Use when creating new skills, editing existing skills, or verifying skills work before deployment
writing-plans
Use when you have a spec or requirements for a multi-step task, before touching code
verification-before-completion
Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always
using-superpowers
Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions
using-git-worktrees
Use when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code