docs-audit
Audit the LaNorme docs for accuracy, currency, no-archeology, clarity, and coverage, then fix what it finds. Use before a release or after changing the CLI, the rules, or any Markdown doc, or when asked whether the docs are solid, current, or up to date. It runs the docs-audit workflow, which checks every documented flag, rule code, config key, and output example against the real tool, flags changelog-style narration outside CHANGELOG and any bluff, and lists undocumented surface.
- License 许可
- MIT
- 所属仓库
- lanorme/lanorme
- 最近更新
- 2026-11-14
- Stars
- ★ 3
- Forks
- 6
01 前言 Front Matter
- name
- docs-audit
- license
- MIT
- metadata
- { "project": "lanorme" }
- description
- Audit the LaNorme docs for accuracy, currency, no-archeology, clarity, and coverage, then fix what it finds. Use before a release or after changing the CLI, the rules, or any Markdown doc, or when asked whether the docs are solid, current, or up to date. It runs the docs-audit workflow, which checks every documented flag, rule code, config key, and output example against the real tool, flags changelog-style narration outside CHANGELOG and any bluff, and lists undocumented surface.
03 SKILL.md 正文
Audit the docs
LaNorme's own prose check (PROSE-001/002/003) already gates em-dashes, British
spelling, and emoji on Markdown. It cannot tell whether a documented flag still
exists or an example still reproduces. This skill fills that gap: it verifies the
docs against the real tool and flags prose that is stale, narrated as history, or
puffed up.
Run the audit
Run the docs-audit workflow from the repo root. Pass the absolute repo path so
the agents work in the right tree:
Workflow({ name: "docs-audit", args: { repo: "<absolute path to the lanorme repo>" } })
If the name does not resolve (for example when the working directory is a different repository), invoke it by path instead:
Workflow({ scriptPath: "<repo>/.claude/workflows/docs-audit.js", args: { repo: "<repo>" } })
Optional args:
docs: an explicit list of Markdown files to audit, instead of letting the workflow discover them.cli: how to invoke the tool if not the dev default (PYTHONPATH=src python3 -m lanorme.cli).
The workflow runs three phases: discover the user-facing docs and snapshot the real CLI surface, audit each doc on two lenses (accuracy and prose) then adversarially verify each finding, and report any user-facing surface that is undocumented. It returns confirmed findings per doc plus a list of coverage gaps.
Act on the result
- Read the returned
auditedfindings andcoverage_gaps. - Fix accuracy issues first: a documented flag, rule code, config key, default, or output example that no longer matches the tool. These are the ones that mislead a reader.
- Remove any changelog-style narration ("previously", "now", "renamed") from docs other than CHANGELOG. State the current behaviour directly. The CHANGELOG is the one place history belongs.
- Cut bluff and redundancy. Keep the prose direct and grounded.
- Close real coverage gaps by documenting the missing flag, format, or key in the doc the workflow suggests.
Verify
After applying fixes, confirm the docs still pass LaNorme's own style gate and that the changes hold up:
lanorme check README.md CONTRIBUTING.md CLAUDE.md docs/RULES.md
Re-run the workflow if you made large changes, and confirm every doc comes back
with verdict solid and no confirmed blocker or major remains.
04 同仓库更多 Skills
valid-minimal
A minimal valid skill used as a parser fixture.
valid-links
A valid skill that links to a bundled reference file.
valid-folded
A valid skill whose description is a folded block scalar spanning multiple lines, to make sure the parser does not call it empty.
unterminated
暂无简介
missing-desc
暂无简介
invalid-no-frontmatter
暂无简介