sure-scale/doc-haus/cite-check
Claude Code and agent skill by sure-scale/doc-haus
- community
Checking that the SKILL.md file is still in the repository.
A procedure for auditing every legal citation in a brief, memo, motion, or opinion letter against the CourtListener database — hallucination detection for legal writing. Use when asked to cite-check, verify citations, validate authorities, confirm cases are real, or review a draft before filing or sending.
Install
npx vskill@latest install sure-scale/doc-haus/cite-checkSKILL.md
Fabricated and mischaracterized citations are the signature failure of AI-assisted legal writing, and courts sanction filers for them. Treat every citation in the document as unverified until a database lookup confirms it. Never mark a citation verified from memory — verification means a live lookup returned the case.
<procedure>
- Extract every citation. Read the full document and list each legal
authority it cites: case citations (full and short-form, including id. and supra references resolved back to their antecedent), statutes, regulations, and rules. For each, record the exact anchor where it appears as [<Document> § <section>] and the proposition the document uses it for (quote the sentence it supports).
- Verify each case against CourtListener. Use the CourtListener MCP
server when available; otherwise use the case-law tool. Search by case name and by reporter citation. Both sources cover U.S. opinions only — for non-U.S. authorities, and for statutes and regulations, do not guess: classify them as unverifiable here and list them for manual confirmation.
- Compare the holding. For each case found, read the returned snippet and
metadata. Confirm the case stands for the proposition the document cites it for — a real case cited for something it does not say is as dangerous as a fake one. Check the cited court and year against the database record.
- Classify each citation into exactly one status:
- verified — the case exists, the reporter citation, court, and year
match, and the retrieved content is consistent with the cited proposition.
- not found — no matching case by name or citation; presumptively
fabricated until the author produces the source.
- ambiguous — a case by that name exists but the citation, court, year,
or party names do not match, or multiple candidates exist, or the source could not verify it (non-U.S. authority, statute, regulation).
- supports-different-proposition — the case is real but the retrieved
content does not support, or contradicts, the proposition cited.
- Output a findings table with one row per citation:
The Evidence column carries the CourtListener URL and matched citation for verified entries, and the specific mismatch for everything else. </procedure>
Close with a summary count by status and a clear verdict: a document is safe to file or send only when every citation is verified or independently confirmed by the supervising attorney. Never soften a not found into "likely correct" — report it as unverified and stop there. Do not invent a replacement citation for a bad one; flag it and let counsel decide.
Security
- Security scanPassed, 96 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 repositoryCheckingdochaus/skill/cite-check/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/sure-scale/doc-haus/cite-check)