superhero-tech/super-brain/lint
Claude Code and agent skill by superhero-tech/super-brain
- community
Checking that the SKILL.md file is still in the repository.
Health check over the compiled wiki in 4-Knowledge/, written up as a ranked report. Use when the human asks to check or lint the wiki, on the monthly maintenance pass, or before the vault gets trusted on a decision that matters.
Install
npx vskill@latest install superhero-tech/super-brain/lintSKILL.md
Lint
Find what has rotted in the wiki, rank it, and hand the list to the human. Fixing is a separate decision.
Write the report in the language the wiki is written in.
How to invoke
- "Check the wiki", "run a lint", "health check"
- The monthly maintenance pass
- Before the human leans on the vault for something expensive
What this catches, and what it does not
Structural problems - missing citations, orphans, index drift, stale pages - it finds reliably, because they are mechanical.
Reasoning errors it finds badly. The model running this check shares its blind spots with the model that wrote the pages, so a clean report means the wiki is tidy, not that it is right. Say this out loud when you hand over a report with no errors.
How it works
1. Decide whether there is anything to check
Count the pages in 4-Knowledge/. Under ten, say there is not enough wiki to lint yet and stop. An empty report trains the human to ignore the next one.
2. Read everything
Read 4-Knowledge/index.md, then every wiki page. This is the one workflow that does not get to work from the index alone.
3. Classify what you find
| Severity | What to look for | |
|---|---|---|
ERROR | Two pages that contradict each other | |
ERROR | Claims with no [source: ...] citation | |
ERROR | Index entries pointing at pages that do not exist, or pages missing from the index | |
ERROR | A sources: entry or [source: ...] citation naming a file that is not in 5-Raw/ | |
WARN | Orphan pages - nothing links to them | |
WARN | Pages marked status: needs_update, or with last_updated older than six months | |
WARN | Pages with source_count: 1 that assert more than one source can carry | |
INFO | Concepts mentioned across several pages with no page of their own | |
INFO | People, companies or tools named on three or more pages with no entity page | |
INFO | Entity pages longer than the concept pages they point at - the argument is in the wrong place | |
INFO | Connections worth making between existing pages |
Every finding names the page and says what to do about it. "Page X has issues" is not a finding.
On the source check: a citation is only checkable if the file it names sits in 5-Raw/. Match on the file name without its extension and compare character for character - a curly apostrophe in the archived file and a straight one in the citation are two different files, and neither looks wrong when you read it on its own. When a source really is absent, say which fix applies: archive the file, or correct the citation. Those are opposite moves and guessing wrong makes it worse.
4. Write the report
Save it to 4-Knowledge/lint-[YYYY-MM-DD].md:
Rank the suggestions by what they unblock, not by severity order. An unranked report gets read once and never again.
5. Log it
Append a lint entry to 4-Knowledge/log.md. That file carries the format.
6. Hand it over
Report the counts and the top suggestion in the conversation, then stop. The human decides what gets fixed.
If they say go ahead, the fixes are a separate pass: make them, append an update entry to 4-Knowledge/log.md saying which pages changed and why, and refresh 4-Knowledge/index.md if any page description moved.
Self-check
- Every finding names a page and a next action
- Every cited source was checked against
5-Raw/, character for character - The report is on disk at
4-Knowledge/lint-[YYYY-MM-DD].md 4-Knowledge/log.mdhas alintentry with the counts- The suggestions are ranked
- A clean report came with the caveat about what this check cannot see
Hard constraints
- Lint reports, it does not repair. Pages change after the human agrees, never during the check.
- Sources in
5-Raw/are read-only. A contradiction between a page and its source is a finding about the page.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryChecking7-Skills/lint/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/superhero-tech/super-brain/lint)