Guide · Agent skill
How to use the Wiki Lint Audit agent skill (wiki-lint)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The wiki-lint skill teaches an AI agent how to check the health of a knowledge wiki that someone else maintains. It reads the wiki's index and pages, looks for seven kinds of problems, and returns a severity-ranked list of findings. This guide explains what agent skills are, how to install this one, how to trigger it, and what it will and will not change. It is written for people who have not used agent skills before.
Wiki Lint Auditpaperclipai/paperclip · MIT Download the skill Quick start
- 1Download the ZIP. Download the wiki-lint ZIP from this page. It contains SKILL.md.
- 2Place the skill folder. Unzip so that wiki-lint/SKILL.md sits in ~/.claude/skills/ for personal use, or in your project's .claude/skills/ folder.
- 3Or upload it in the Claude apps. In the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- 4Confirm the requirements. Check that the Paperclip LLM Wiki plugin tools are available and that you have an operation issue with operationType lint, plus its wikiId and spaceSlug.
- 5Start a lint run. Create or open a lint operation issue, or ask the agent for a health check of the named wiki and space.
- 6Triage the findings. Review the severity-grouped list, check each evidence quote, and decide which follow-up operation to run.
- 7Check the log. Confirm the run's entry was appended to wiki/log.md, which is the only wiki file the skill may change.
What an agent skill is and what wiki-lint makes the agent do
An agent skill is a folder containing a SKILL.md file. The file gives an AI agent instructions for one specific kind of task, and the agent loads those instructions when a request matches the skill's description. This skill's description says it applies when an LLM Wiki operation issue is a lint or health check.
Once loaded, the agent audits a wiki for seven recurring issues, checked in this order: contradictions, stale claims, orphan pages, concept gaps, broken [[wiki-links]], weak provenance, and index/log drift. It then returns a triage list grouped as critical, medium and low, with file paths, a short quote as evidence, a suggested fix, and the follow-up operation (ingest, paperclip-distill, index-refresh or manual review).
- Contradictions: two pages make incompatible claims about the same entity, decision or status.
- Stale claims: a page asserts something a newer source under raw/ has superseded.
- Orphan pages: a wiki page is not linked from index.md and not referenced by any other wiki page.
- Concept gaps: a term appears on three or more pages but has no dedicated wiki/concepts/<slug>.md page.
- Broken links: a [[wiki-link]] points to a file that does not exist.
- Weak provenance: a non-trivial claim is uncited or only cites the wiki itself.
- Index/log drift: pages missing from index.md, index entries for deleted pages, or log entries with no matching page change.
Install the wiki-lint skill
Download the ZIP for this skill from this page. The package contains a single file, SKILL.md. Unzip it so that the folder named wiki-lint sits inside one of the following locations.
For a personal install that applies to all your projects, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: ~/.claude/skills/wiki-lint/SKILL.md
- Project install: <your-project>/.claude/skills/wiki-lint/SKILL.md
- Claude apps: Settings → Capabilities → Skills → upload
~/.claude/skills/wiki-lint/SKILL.mdRequirements: what you need before it runs
This skill is built for the Paperclip platform with the LLM Wiki plugin installed. It relies on specific tools from that plugin: wiki_search, wiki_read_page, wiki_list_sources, wiki_read_source and wiki_write_page. The agent can only lint if those tools are available to it.
The skill also expects a specific input. You need an operation issue whose operationType is lint, along with its wikiId and spaceSlug. The wiki must use the Paperclip LLM Wiki layout, with wiki/ and raw/ folders, an index.md, and a log.md. The skill lints only the target space unless the issue explicitly asks for a multi-space sweep.
- Paperclip platform with the LLM Wiki plugin and its wiki tools
- An operation issue with operationType set to lint
- The issue's wikiId and spaceSlug
- A wiki laid out with wiki/, raw/, index.md and log.md
Trigger the skill with natural requests
You do not need a special command. The agent uses the skill when your request matches its purpose. The most reliable trigger is a lint operation issue in the Paperclip platform, because the skill's inputs come from that issue. In a plain chat, ask for a health check of the wiki and name the wiki and space so the agent can pass the right identifiers to its tools.
Phrasing that matches the skill's description works well, such as asking for a lint or health check, an audit for contradictions or orphan pages, or a check of index and log drift.
- "Run a lint health check on wiki W, space marketing-docs."
- "Audit this LLM Wiki for contradictions and broken links, and give me a triage list."
- "Check the index and log for drift in the operations space."
What you get back and what stays untouched
The output is a findings list, not a set of repairs. Each finding is one bullet with its file path, evidence, suggested fix and follow-up operation. The report leads with the count by severity. Where the agent is unsure how serious an issue is, the skill tells it to rate the item medium and add a verify note.
The skill is strict about writes. It must not change files under raw/, and it must not change files under wiki/ except wiki/log.md. At the end of each run it appends a short entry to the log with the date, the number of findings by severity, and the operation issue identifier. If a run finds nothing, the issue is closed with 'no findings' and the log entry still records that the audit happened.
- Severity groups: critical (contradictions, broken links to active pages, fabricated citations), medium (stale claims, weak provenance, large concept gaps), low (orphans, log drift, small index gaps)
- Only write permitted to the wiki: a log entry in wiki/log.md
- Fixes are carried out separately by you or by the follow-up operation
Safety notes: scripts and network
This skill does not run scripts and does not use the network. Its instructions are limited to reading wiki content through the plugin tools and appending one log entry. Because it cannot edit pages, the main risk is an incorrect finding, so check the evidence quotes before acting on a critical item.
The skill also does not check claims against the outside world. It only compares pages with sources inside the same wiki. A claim can pass lint and still be false.
- No scripts are included or run
- No network access is used
- Findings are based only on wiki pages and sources in the wiki
Tips for getting useful audits
Run lint on a schedule or after a batch of ingests so that new problems are caught while they are small. Keep your index.md and log.md current, because index and log drift checks depend on them. Treat each finding as a suggestion: decide which follow-up operation fits, then run it yourself.
If you have several spaces, lint each one with its own operation issue rather than asking for a sweep, since the skill is designed to stay inside one space by default.
- Lint after batches of ingests to catch problems early
- Keep index.md and log.md up to date
- Act on critical findings first, then medium, then low
- Use one operation issue per space
What people use it for
- Periodic health checks of an LLM Wiki maintained inside the Paperclip platform.
- Triaging wiki quality before deciding what to ingest or refresh next.
- Finding pages that nobody links to and deciding whether to link, merge or remove them.
- Spotting [[wiki-links]] that point to missing files before readers hit them.
- Using the skill as a read-only reference pattern when building your own knowledge-base audit workflow.
Frequently asked questions
How do I install the wiki-lint agent skill?
Download the ZIP from this page and unzip it so that wiki-lint/SKILL.md sits in ~/.claude/skills/ for personal use or in your project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
What do I need to run wiki-lint?
You need the Paperclip platform with the LLM Wiki plugin, which provides wiki_search, wiki_read_page, wiki_list_sources, wiki_read_source and wiki_write_page. You also need an operation issue with operationType set to lint, plus its wikiId and spaceSlug.
Does wiki-lint fix problems automatically?
No. The skill is designed to audit only. It must not change raw sources or wiki pages, apart from appending a short entry to wiki/log.md. Fixes are carried out separately.
Is wiki-lint free and what is its licence?
The source for this skill is published under the MIT licence. The source does not mention pricing, so check the project's repository for any other terms.
Can wiki-lint check whether claims are true in the real world?
No. It only compares pages with sources inside the same wiki, so a claim can pass lint and still be wrong in the outside world.
Does wiki-lint work with wikis outside Paperclip?
It is built for the Paperclip LLM Wiki format, which depends on wiki/ and raw/ folders, index.md, log.md and specific wiki tools. Other systems may not provide those, so the skill may not work there without adaptation.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use wiki-lint if you maintain an LLM Wiki in the Paperclip platform and want a read-only, severity-ranked health check that you act on yourself. Pick something else if you want automatic repair, if your wiki uses a different layout or tools, or if you need claims checked against the outside world.
Download the skill