Description

Say 'lint my docs' to scan your markdown KB for broken links, orphan pages, and missing metadata — then auto-fix them.

README (SKILL.md)

Doc Doctor

Name: Claude Doc Doctor
Author: singggggyee

Your markdown docs have broken links, orphan pages, missing metadata, and you don't know it. This skill finds them all and fixes them for you.

Quick start

Just say:

"lint my docs" — scan your project's markdown files
"check my KB" — get a health score out of 100
"fix my wiki" — auto-fix broken links, missing frontmatter, bad filenames
"lint my memory" — audit your Claude Code memory files

What it checks

Check	What it finds	Auto-fixable?
Broken links	`[[wiki-links]]` pointing to nothing	Yes — fuzzy-matches the right target
Orphan pages	Files no other page links to	Suggests where to add links
Missing frontmatter	No title, tags, or metadata	Yes — generates from content
Thin articles	Under 100 words	Suggests outlines
Bad filenames	Spaces, non-kebab-case	Yes — renames + updates all references
Inconsistency	Mixed tag casing, date formats	Yes — normalizes

Links

GitHub: SingggggYee/kb-lint
PyPI: pip install kb-lint

Below are instructions for Claude Code. You can ignore this section.

Instructions

Find the target. User-specified path, or scan for docs/, wiki/, notes/, content/. If user says "memory", target ~/.claude/.
Run the linter.
```
kb-lint --version 2>/dev/null || pip install kb-lint
kb-lint \x3Cpath> --format json --severity info 2>&1
```
If kb-lint unavailable, check files manually for: broken [[wiki-links]], missing frontmatter, thin articles (\x3C100 words), orphans, bad filenames.
Show a health dashboard. Score out of 100, issues grouped by category and severity, most critical first.
Offer to fix. Go beyond the CLI:
- Broken links: fuzzy-match the intended target, confirm with user
- Frontmatter: generate meaningful title/tags/confidence from content, get created date from git history
- Orphans: suggest which articles should link to the orphan
- Thin articles: suggest an outline, never fabricate content
- Filenames: rename to kebab-case, update all references
Verify. Re-run linter, show before/after score.

Claude Code memory mode: For ~/.claude/ targets, also check MEMORY.md index accuracy, validate memory frontmatter, flag stale entries (>90 days), find duplicates.

Rules: Never fabricate content. Always show changes before applying. Respect \x3C!-- kb-lint-disable --> inline comments and per-check severity overrides.

Usage Guidance

This skill appears to do what it says, but it will read and modify files in your repo (and optionally your Claude memory at ~/.claude/), and may install a Python package. Before installing or running: 1) Review the kb-lint project (homepage/PyPI source) to confirm you trust it. 2) Prefer running pip install inside a virtual environment (or inspect the package) instead of global install. 3) Back up or use git so you can review/undo changes; run in 'show changes / dry-run' mode first. 4) Be cautious when scanning your home directory (~/.claude/) because it can expose personal data—only run memory-mode if you intend it. 5) Note the small inconsistency between the 'uv' install spec and the SKILL.md's pip fallback; confirm how your platform will install the kb-lint package. If you want extra assurance, run kb-lint manually on a test copy of your docs before granting the skill write permission.

Capability Assessment

✓ Purpose & Capability

The name/description match the requested binaries and actions: python3 is required and the skill installs/uses a Python CLI (kb-lint) to lint and auto-fix markdown. No unrelated credentials, config paths, or network-only services are requested. One minor inconsistency: the install spec lists kind: uv package: kb-lint while the runtime instructions use pip as a fallback; this is explainable (platform installer vs. pip) but worth noting.

ℹ Instruction Scope

SKILL.md explicitly instructs scanning project folders (docs/, wiki/, etc.) and, in 'memory' mode, ~/.claude/. It runs kb-lint CLI and, if missing, runs pip install kb-lint. The skill will rename files, update references, read git history to infer creation dates, and propose or perform auto-fixes. Those actions are within scope for an auto-fixer, but they are materially destructive (renames/edits). The doc includes safety rules (always show changes before applying, confirm fuzzy matches, never fabricate content), which mitigates risk if followed. Users should expect file I/O and git access.

ℹ Install Mechanism

Install spec uses an 'uv' package installer for kb-lint; the runtime fallback installs via pip. Installing a third-party Python package from PyPI is a common and expected mechanism for a Python CLI, but it does execute package installation on the host (network fetch, files written to disk, potential post-install scripts). There are no direct downloads from unknown/personal URLs in the spec.

✓ Credentials

The skill requests no environment variables or credentials. It does access user files and git history when run (including optional ~/.claude/), which is proportional to a linter that can operate on local KB/memory. No unexplained secrets or broad service credentials are requested.

✓ Persistence & Privilege

always is false and the skill does not request persistent platform privileges. It installs a package into the host environment when run (per install spec / pip fallback), but it does not try to modify other skills or global agent config per the provided instructions.

Version History

v4.0.1

Rename to Claude Doc Doctor

v4.0.0

Fix display name

v3.0.0

Fix display name to Doc Doctor

v1.0.0

Initial release

v2.0.0

Initial release

Metadata

Slug doc-doctor

Version 4.0.1

License MIT-0

All-time Installs 0

Active Installs 0

Total Versions 5

Frequently Asked Questions

What is Claude Doc Doctor?

Say 'lint my docs' to scan your markdown KB for broken links, orphan pages, and missing metadata — then auto-fix them. It is an AI Agent Skill for Claude Code / OpenClaw, with 149 downloads so far.

How do I install Claude Doc Doctor?

Run "/install doc-doctor" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Claude Doc Doctor free?

Yes, Claude Doc Doctor is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Claude Doc Doctor support?

Claude Doc Doctor is cross-platform and runs anywhere OpenClaw / Claude Code is available (macos, linux).

Who created Claude Doc Doctor?

It is built and maintained by SingYee (@singggggyee); the current version is v4.0.1.

More Skills

Claude Doc Doctor