Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →This skill should be used when the user wants to review, audit, or sanity-check an existing agent skill in this repo against authoring conventions. Common triggers include "review the X skill", "audit this skill", "check skill X against repo conventions", "is this skill any good", "second opinion on skill X", "sanity check skill X", and "review skill X before publishing". Produces a severity-tiered report (error / warn / info) with a Clean section even on pass, and classifies the skill type. Ski
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 197% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 148% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 81% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-15 | ✓→✗ | ▼ Worse | 39% | 0% |
Reviews an existing skill in this repo against authoring conventions. Produces a severity-tiered report (error / warn / info) plus a Clean section, and classifies the skill type.
$ARGUMENTS — one of:
ts-best-practices) — looks under skills/<name>/ then .agents/skills/<name>/skills/foo/SKILL.md or skills/foo)Locate the skill directory. Confirm SKILL.md exists. Run:
bashpnpm skill-toolkit lint <skill-name>
Capture the lint output verbatim — it's the floor, not the ceiling. Lint passing means mechanical rules pass; it does not mean the skill is well-authored.
Pick exactly one — this dictates which audit lens to apply:
| Type | Examples | Audit focus | | -------------- | ------------------------------------------------------------- | ------------------------------------------------------------------------------------ | | Discipline | "always run the test", "never use any", "always use Result" | Rationalization table present? Body covers the realistic rationalizations? | | Technique | "use ts-pattern for branching", "use zod for parsing" | Triggers cover phrasings? Workflow is concrete (not abstract)? | | Pattern | "use \Params for ≥2-arg fns", "kebab-case files" | Description's `Skip when` covers counter-examples + recognition (when NOT to fire)? | | Reference | "API X works like…", "convention Y says…" | Declines questions outside its scope? Reference depth matches the surface it claims? |
State the classification explicitly. If you can't classify it cleanly, that's itself a finding (the skill's purpose is fuzzy).
Lint enforces frontmatter shape, naming, anti-shortcut words. Deep references cover what lint can't:
skill-creator/references/description.md — description quality beyond char countskill-creator/references/frontmatter.md — frontmatter schemaskill-creator/references/xml-usage.md — when to use <example> / <good> / <bad>If you skip these and only cite lint-checklist.md, you're guessing at depth.
Beyond the lint pass:
Skip when … clause naming what the skill does NOT do## When to use / ## When NOT to use sections — routing is the description's job, and dispatchers never read the body anyway; duplicating triggers wastes tokens and creates drift risk (warn if present)argument-hint, user-invocable, optionally disable-model-invocation and allowed-tools) and fenced behind the # --- Claude Code extensions comment. Flag any use of fabricated fields (model-invocable, metadata) — these aren't in the official Claude Code spec; see skill-creator/references/frontmatter.md.## sections<example> blockTODO / FIXME / XXX## Rationalization table section (per skill-creator step 6) — its absence on a discipline skill is a warnUse this exact format. The Clean section is mandatory — it forces grounding the verdict in specific rules rather than vibes.
SUMMARY: <N> findings (<E> error / <W> warn / <I> info) — <skill-type> skill
## ERRORS
✗ <file>:<line> <one-line-finding>
<body — what's wrong, why it matters, suggested edit>
## WARNS
⚠ <file>:<line> <one-line-finding>
<body>
## INFOS
ℹ <file>:<line> <one-line-finding>
<body>
## Clean
- <specific rule that passes — e.g., "Description has 5 verbatim triggers in double quotes (description.md:38)">
- <another specific rule>
- <`pnpm skill-toolkit lint <name>` — <pass/fail summary>>Empty severity sections render as ## ERRORS\n(none) — keeps the shape comparable across runs. A pass verdict is 0 error / 0 warn / 0 info plus a populated Clean section.
<bad> "Looks solid, ship it." </bad>
No grounding. The pass verdict carries no information the user can act on or revisit later. Use a Clean section listing specific rules.
<bad> "Lint passes, looks fine." </bad>
Lint is the floor. This is the lazy-reviewer failure mode — see step 3.
<bad> Manufactured nits because the skill seemed too clean and the reviewer wanted to look thorough. </bad>
If a skill genuinely has 0 findings, say so in the Clean section. Inventing critique to look thorough is worse than missing real issues.
<example> <input>"review skills/ts-best-practices — does it still hold up against our conventions?"</input> <output>
skills/ts-best-practices/. Run pnpm skill-toolkit lint ts-best-practices → clean.description.md, frontmatter.md, xml-usage.md.SKILL.md has 6 triggers, description and README.md have 5. Description missing "audit this ts file".@pkg/fp in conditionals table leaks an internal-monorepo assumption.SUMMARY: 3 findings (0 error / 2 warn / 1 info) — discipline skill
## ERRORS
(none)
## WARNS
⚠ description (SKILL.md:5) trigger parity drift vs. README.md (5 vs 6 triggers; missing "audit this ts file")
Add the missing trigger to description and README — the description is the dispatcher's surface.
⚠ SKILL.md:184 `@pkg/fp` reference leaks a private-monorepo assumption into a public skill
Replace with "ternary helper" or add a one-line reference link.
## INFOS
ℹ description (SKILL.md:11) Skip-when bundles two cases with an em dash
Split the framework case (React/Vue/Svelte) from the functional-refactor redirect.
## Clean
- Description has Use-when phrase + 5 verbatim triggers in double quotes (description.md:38)
- Frontmatter has the documented Claude Code extension fields (argument-hint, user-invocable)
- Body has 2 <example> blocks (xml-usage.md ≥ 1 required)
- Rationalization table present — appropriate for a discipline skill (skill-creator step 6)
- pnpm skill-toolkit lint ts-best-practices — 0 error / 0 warn / 0 info</output> </example>
<example> <input>"sanity-check skills/skill-portability — i think it's fine"</input> <output> Lint clean. Classified as a technique skill (replaces ad-hoc provider knowledge with a structured providers.mjs lookup).
SUMMARY: 0 findings (0 error / 0 warn / 0 info) — technique skill
## ERRORS
(none)
## WARNS
(none)
## INFOS
(none)
## Clean
- Description: 720 chars, has Use-when, 5 verbatim triggers, explicit Skip-when (description.md ✓)
- Frontmatter: argument-hint, user-invocable set; no fabricated fields
- Body: ≥ 3 `## ` sections, ≥ 1 `<example>` block, no TODO/FIXME
- Companions: README.md, LICENSE, references/, scripts/ all present
- Single-source-of-truth in providers.mjs reflects the skill's own thesis
- pnpm skill-toolkit lint skill-portability — 0/0/0Nothing manufactured. If you want depth beyond the structural review, run node skills/skill-portability/scripts/providers.mjs --check to confirm the docUrls are still 200 — that's the skill's own staleness check.
</output> </example>
Captured from baseline transcripts where reviewers without this skill skipped rules. Future reviewers: recognize your own pattern.
| Skipped rule | Verbatim excuse | Why it's wrong | | ---------------------------------------------------------------------- | ------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Read deep references beyond lint-checklist.md | "relied on lint-checklist summary" | Lint enforces mechanical rules; the deep references cover description quality, frontmatter shape, and XML usage rules that lint cannot check | | Classify the skill type (discipline / technique / pattern / reference) | (omitted entirely) | Different types need different audits — discipline skills require a rationalization table; pattern skills need recognition tests; without classification you're applying the wrong lens | | Use severity-tiered output (error / warn / info) even on a pass | "used numbered findings" / "prose verdict" | Comparable output across runs; numbered lists drift in shape; prose ("looks solid") invites manufactured-nits or vague-pass failure modes | | Include a Clean section listing what specifically passes | "said 'looks solid' / 'ship it'" | Pass verdicts without specifics rot — six months later nobody knows what was actually checked. Clean sections force grounding in specific rules |
skill-creator/SKILL.md — authoring workflow this reviewer audits againstskill-creator/references/description.md — description quality rulesskill-creator/references/frontmatter.md — frontmatter schemaskill-creator/references/lint-checklist.md — mechanical rules (the floor)skill-creator/references/xml-usage.md — <example> / <good> / <bad> boundariescode-reviewer/references/review-output-format.md — three-tier output spec inspirationOther measured skills in the registry, with their headline benchmark lift.