Agent skill
kb-lint
Health check and lint a knowledge base wiki for consistency and gaps.
Install this agent skill to your Project
npx add-skill https://github.com/notque/claude-code-toolkit/tree/main/skills/kb-lint
SKILL.md
KB Lint Skill
Health check a knowledge base wiki under research/{topic}/ for structural consistency, broken references, and content quality gaps. Produces a structured lint report.
Invocation
/kb-lint --topic TOPIC [--fix]
--topic TOPIC— the topic slug underresearch/--fix— attempt to auto-fix structural errors (missing index entries, broken links)
Phase 1: INVENTORY
Read research/{topic}/kb.yaml. Count all files:
raw/— raw source markdown fileswiki/concepts/— concept articleswiki/sources/— source summary articleswiki/queries/— saved query outputs
Read 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"}.
Phase 2: STRUCTURAL CHECKS
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:
- Collect all failures with specific file names
- Accumulate counts: errors, warnings
Phase 3: CONTENT QUALITY
LLM-judged quality checks on concept articles. Read up to 10 concept articles (or all if fewer than 10). For each:
Article-level checks:
- Is the article just a title with no substance? (fewer than 3 sentences of actual content)
- Does the article have a definition or overview section?
- Does the article contain contradictory statements?
Cross-article checks (after reading the sample):
- Concept overlap: are two concepts covering substantially the same topic? Flag pairs and suggest merging.
- Gap analysis: based on sources referenced and concepts present, are there obvious missing concepts? List up to 5 significant gaps only — do not suggest minor terms.
Phase 4: REPORT
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.
Auto-fix behavior (--fix)
If --fix was specified AND structural errors were found, offer to repair:
- Missing index entries: append missing file paths to
_index.mdarticles list - Broken
[[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.
Anti-patterns
- Do NOT fix issues silently — always report before fixing
- Do NOT suggest concepts for every minor term — only significant gaps (max 5)
- Do NOT report passing checks — only failures and suggestions
- Do NOT read every raw source during lint — structural checks use wiki files only
- Do NOT treat missing
research/{topic}/as a lint failure — exit with a clear error message before phase 1
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
voice-writer
Unified voice content generation pipeline with mandatory validation and joy-check. 9-phase pipeline: LOAD, GROUND, GENERATE, VALIDATE, REFINE, JOY-CHECK, 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".
image-auditor
Non-destructive image validation for accessibility and health.
video-editing
Video editing pipeline: cut footage, assemble clips via FFmpeg and Remotion.
comment-quality
Review and fix temporal references in code comments.
e2e-testing
Playwright-based end-to-end testing workflow.
anti-ai-editor
Remove AI-sounding patterns from content.
Didn't find tool you were looking for?