hai-audit-docs-internally
$
npx mdskill add hylarucoder/hai-stack/hai-audit-docs-internallyAudits documents for internal consistency and produces a prioritized repair plan
- Identifies contradictions, duplication, and stale content in documentation
- Analyzes structure, terminology, and claims without comparing to code
- Prioritizes findings by severity and suggests update/move/merge/remove/split actions
- Delivers a clear report with repair order and decisions for each issue
SKILL.md
.github/skills/hai-audit-docs-internallyView on GitHub ↗
--- name: hai-audit-docs-internally description: | Audits a document or doc set from the inside and produces a prioritized findings report (P0-P3) of internal conflicts, stale content, terminology drift, duplication, and misplaced or unsupported claims, plus update/move/merge/remove/split decisions and a suggested repair order. Use this whenever the user wants docs sanity-checked or cleaned up for self-consistency without comparing against code, even when they never say "audit": contradictory or duplicate sections, stale assumptions, unclear structure, or a PRD/spec that should hang together. Trigger on casual and Chinese phrasings too: 文档自相矛盾, 文档前后不一致, 文档内部冲突, 审一下这份文档, 帮我审审这个 PRD, 文档体检, "these two sections disagree", "is this spec consistent", "the docs repeat themselves". If the truth source is the code, use hai-audit-docs-against-code instead. --- # Hai Audit Docs Internally For Chinese readers, see `SKILL.zh_CN.md`. The English `SKILL.md` is the execution source of truth. ## Overview Audit one document or a documentation set from the inside, then deliver a prioritized findings report with repair decisions. The source of truth is the document's own goal, structure, terminology, claims, and cross-references — not the codebase. ## Core Principle A document is a coherent argument, and the audit reads it as one. Judge it for consistency of purpose, terms, and claims: every section should serve the same target, the same concept should carry the same name, and nothing should assert what another section denies or what no evidence supports. The enumerated consistency checks live in Workflow step 3. ## Workflow 1. Identify the document scope. - Single document, related docs set, PRD pack, README plus docs, or generated spec set. - Note the document's apparent audience, purpose, and expected decision/use. 2. Build a document map. - List the main sections and what each section is trying to do. - Extract key claims, definitions, terminology, assumptions, scope boundaries, examples, dates, owners, statuses, and cross-links. - Identify repeated claims and places where the same concept appears under different names. 3. Find internal issues. - **Direct conflict**: two sections make incompatible claims. - **Scope conflict**: in-scope and out-of-scope sections disagree. - **Terminology drift**: the same concept uses different names, or one name means different things. - **Lifecycle conflict**: statuses, phases, dates, or dependencies do not line up. - **Acceptance conflict**: success criteria do not prove the stated goal. - **Stale signal**: text references old decisions, old names, old dates, deprecated sections, or superseded assumptions. - **Redundant content**: repeated paragraphs, examples, or checklists should be merged or removed. - **Misplaced content**: implementation detail, policy, background, or task planning lives in the wrong document section. - **Unsupported claim**: a strong claim lacks evidence, owner, source, or decision record. 4. Decide the repair type. - **Update** when the content is useful but stale or imprecise. - **Move** when the content belongs elsewhere in the same document. - **Merge** when repeated content fragments one idea. - **Remove** when content is out of scope, obsolete, unsupported, or harmful. - **Split** when one document contains multiple independent goals or audiences. - **Ask** when a conflict cannot be resolved from the document itself. 5. Produce the audit. - Deliver: Verdict, Document Map, prioritized Findings (each with type, location, evidence, impact, repair), a Remove/Update/Move decision table, Open Decisions, and a Suggested Repair Order. - Read `references/output-template.md` before finalizing. - Prioritize issues that change understanding, decisions, scope, or execution. - Keep wording fixes secondary unless wording causes ambiguity or conflict. ## Severity | Level | Meaning | |-------|---------| | P0 | Internal conflict could cause a wrong decision, unsafe action, or failed launch | | P1 | Core goal, scope, terminology, or acceptance is inconsistent | | P2 | Stale, duplicated, misplaced, or unsupported content creates confusion | | P3 | Minor clarity, structure, or formatting issue | | Needs decision | The document has a real fork that requires owner input | ## Use a different skill when - The truth source is the code: use `hai-audit-docs-against-code` to check docs against implementation, config, and API contracts. This skill never compares against code. - The document is a PRD that needs product-requirement crafting or repair (scope, acceptance, structure as a spec): use `hai-prd`. - The docs just need formatting and presentation cleanup, not consistency judgment: use `readme-beautifier`.
More from hylarucoder/hai-stack
- ast-grep-rule-crafter|
- clean-code-reviewerProduces a severity-rated (高/中/低) Clean Code findings report across 7 dimensions (naming, function size/SRP, duplication/DRY, over-engineering/YAGNI, magic numbers, structural clarity, project conventions), each with a location and a behavior-preserving refactor suggestion — never changing functionality. Use whenever the user asks for a code review, quality check, refactor advice, or code-smell / Clean Code analysis, OR points at a file/function/diff and asks if it is well-written, too long, too repetitive, over-engineered, or poorly named — even casually, and even if they never say "review" ("I just wrote this, look it over", "does this look good before I commit"). Trigger on 代码体检, 代码质量, 重构检查, 代码审查, 这段代码写得怎么样, 帮我看看代码有没有问题, 有没有坏味道, 这函数是不是太长了, 命名规范吗, 魔法数字, 重复代码, 过度设计, and English like "is this code clean", "any code smells", "check this file".
- create-visual-card|
- entity-model-auditor|
- geju|
- goudi|
- hai-architecture|
- hai-audit-docs-against-code|
- hai-goal|
- hai-idea|