skills/kb-lint/SKILL.md
Health check and lint a knowledge base wiki for consistency and gaps.
npx skillsauth add notque/claude-code-toolkit kb-lintInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Health check a knowledge base wiki under research/{topic}/ for structural consistency, broken references, and content quality gaps. Produces a structured lint report.
/kb-lint --topic TOPIC [--fix]
--topic TOPIC — the topic slug under research/--fix — attempt to auto-fix structural errors (missing index entries, broken links)Read research/{topic}/kb.yaml. Count all files:
raw/ — raw source markdown fileswiki/concepts/ — concept articleswiki/sources/ — source summary articleswiki/queries/ — saved query outputsRead research/{topic}/wiki/_index.md.
Report current state:
KB '{topic}': N raw sources, M concepts, P source summaries, Q queries.
Last compiled: {last_compiled from _index.md frontmatter, or "never"}.
Run deterministic checks. Read file contents and frontmatter as needed. Do NOT read raw sources for structural checks — wiki files only.
| Check | Pass Condition | Severity |
|-------|----------------|----------|
| Raw coverage | Every raw/*.md has a matching wiki/sources/*.md | ERROR |
| Concept backlinks | Every concept slug in source frontmatter concepts: list exists as wiki/concepts/{slug}.md | ERROR |
| Source backlinks | Every source slug in concept frontmatter sources: list exists as wiki/sources/{slug}.md | WARN |
| Index completeness | Every .md file under wiki/ (excluding _index.md) appears in _index.md | ERROR |
| Orphan concepts | Concepts that appear in 0 source concepts: lists | WARN |
| Empty articles | Wiki files with fewer than 50 words of content (excluding frontmatter block) | WARN |
| Stale index | _index.md frontmatter last_compiled is more than 7 days old | WARN |
| Missing frontmatter | Any wiki file (concepts, sources, queries) that lacks a valid YAML frontmatter block | ERROR |
| Broken cross-links | Markdown links [text](path) within wiki files that point to non-existent files | ERROR |
For each check:
LLM-judged quality checks on concept articles. Read up to 10 concept articles (or all if fewer than 10). For each:
Article-level checks:
Cross-article checks (after reading the sample):
Produce a structured report to stdout. Format:
# KB Lint Report: {topic}
## Summary
- Total: N raw, M concepts, P sources, Q queries
- Errors: X | Warnings: Y | Suggestions: Z
## Errors
- [ERROR] {description with specific file name(s)}
## Warnings
- [WARN] {description with specific file name(s)}
## Suggestions
- [SUGGEST] {description}
## Recommended Actions
1. {specific action to fix most critical issue}
2. {next action}
3. ...
If there are no errors, no warnings, and no suggestions, output:
KB '{topic}' is healthy. No issues found.
--fix)If --fix was specified AND structural errors were found, offer to repair:
_index.md articles list[[wiki/path]]-style links: report only (cannot safely infer correct target)Before fixing, print each planned change and ask for confirmation. Do NOT auto-fix content quality issues — those require human or LLM judgment during compile.
research/{topic}/ as a lint failure — exit with a clear error message before phase 1documentation
Document translation: quick/normal/refined modes with chunked parallel subagents and glossary support.
development
AI image generation: Gemini and Nano Banana backends; single/series/batch workflows with prompt-to-disk.
testing
Unified voice content generation pipeline with mandatory validation and joy-check. 13-phase pipeline: LOAD, GROUND, STATS-CHECKPOINT, GENERATE, HOOK-GATE, VALIDATE, REFINE, VARIETY-GATE, JOY-CHECK, ANTI-AI, CLOSE-GATE, OUTPUT, CLEANUP. Use when writing articles, blog posts, or any content that uses a voice profile. Use for "write article", "blog post", "write in voice", "generate content", "draft article", "write about".
documentation
Critique-and-rewrite loop for voice fidelity validation.