From docs-guardian
Defines standards for evaluating documentation quality: severity tags, finding formats, coverage/staleness metrics, and quality scores. Use for structured doc audits.
How this skill is triggered — by the user, by Claude, or both
Slash command
/docs-guardian:standardsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
| Tag | Meaning | When to use |
| Tag | Meaning | When to use |
|---|---|---|
CRITICAL | Documentation is dangerously wrong — will mislead users into bugs or security issues | API signature changed but docs show old signature; security-relevant parameter undocumented |
HIGH | Documentation is missing or significantly stale — users will struggle | Public function has no docs; doc references deleted parameter |
MEDIUM | Documentation exists but is incomplete or outdated in non-critical ways | Missing examples; parameter description vague; minor version drift |
LOW | Documentation quality issue — cosmetic or style | Typos; inconsistent heading levels; missing period at end of sentence |
INFO | Observation, not a problem | Unusually thorough docs; suggestion for improvement |
Every finding reported by an agent MUST use this structure:
### [SEVERITY] Short title
- **File**: `path/to/file`
- **Line**: 42 (if applicable)
- **Related doc**: `docs/path/to/doc.md` (if applicable)
- **Detail**: One-paragraph explanation of the issue.
- **Suggestion**: Concrete action to fix it.
Every agent output MUST begin with:
## <Agent Name> Report
**Scanned**: <number> files
**Findings**: <number> (Critical: N, High: N, Medium: N, Low: N, Info: N)
**Timestamp**: <ISO 8601>
Coverage is reported as a percentage:
coverage% = (documented_public_symbols / total_public_symbols) × 100
A "documented" symbol has at minimum:
Staleness is measured by comparing:
code_last_modified: last git commit touching the source filedoc_last_modified: last git commit touching the corresponding doc fileA doc is stale when code_last_modified - doc_last_modified > stalenessThresholdDays (from config).
Quality is rated 0–100 based on:
| Factor | Weight | Criteria |
|---|---|---|
| Completeness | 30% | All public APIs documented |
| Accuracy | 30% | Docs match current code |
| Freshness | 20% | No stale docs beyond threshold |
| Readability | 10% | Clear language, good structure |
| Examples | 10% | Working code examples present |
npx claudepluginhub xiaolai/claude-plugin-marketplace --plugin docs-guardianRuns an interview-style session to sharpen a plan or design, producing ADRs and a glossary as you go.
Quick-reference card listing all ponytail modes (Lite, Full, Ultra), skills, and commands. Useful for discovering or recalling ponytail capabilities.