source-verify
This skill should be used when the user asks to 'verify sources', 'check citations', 'fact-check footnotes', 'verify quotes', 'source check', 'are my cites real', 'check for hallucinated citations', or needs to verify that citations in a legal/academic manuscript actually exist and are accurate. Also use when the user wants to check that quoted text actually appears in the cited source.
npx skills add majiayu000/claude-skill-registry --skill source-verify --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Verify that citations in a manuscript are real, accurate, and that quoted text actually appears in the source. Operates as a domain-specific audit-fix-loop: extract citations, run checks, score, fix, re-check.
Announce: "Using source-verify to check citations against Paperpile and source documents."
How it works
Performs four checks in sequence:
- Check 1: EXISTENCE (mechanical): grep paperpile.bib for author + title + year to confirm existence; if not found, flag as UNVERIFIED.
- Check 2: ACCURACY (mechanical): compare volume, issue, pages, year against bibtex; flag any mismatches as FIELD_ERROR.
- Check 3: QUOTE VERIFICATION (outsourced to RAG): if the footnote contains a direct quote, verify the quoted text appears in the source using Readwise highlights, or rga against a downloaded PDF, or NLM generate-chat as fallback; flag QUOTE_NOT_FOUND or QUOTE_MISMATCH.
- Check 4: CLAIM GROUNDING (outsourced to RAG): verify whether the cited source supports the claimed statement using Readwise/NLM chat; flag UNSUPPORTED or CONTRADICTED.
Checks 1-2 run on every invocation; Check 3 runs only when a direct quote is present; Check 4 runs only when explicitly requested by the user.
Mechanical checks rely on paperpile.bib, Drive PDFs, and Readwise highlights; LLM-based checks are used only as described when needed. The tool never uses memory as ground truth for verification.
When to use it
Use when the user needs to verify that citations exist, fields match, quotes appear in the source, or, optionally, to ground claims to sources. Triggered workflow differs by whether the manuscript contains quotes and whether the user requests deep claim grounding.
What it can touch
Requires access to:
- paperpile.bib on Drive
- manuscript footnotes or extracted footnotes in /tmp
- Drive PDFs for quotes verification
- Readwise highlights for quotes verification
- Optional: NLM notebook for deeper grounding
Caveats
License: MIT. Limits include that Statutes are not ground-truth verifiable; the tool flags SKIPPED_NO_GROUND_TRUTH for such cases. Some checks depend on external services (Readwise, Drive) and may require authentication. Mechanical checks are deterministic and precede any LLM-based verification.
# Source Verification Verify that citations in a manuscript are real, accurate, and that quoted text actually appears in the source. Operates as a domain-specific audit-fix-loop: extract citations, run checks, score, fix, re-check. **Announce:** "Using source-verify to check citations against Paperpile and source documents." ## What This Skill Checks ``` ┌──────────────────────────────────────────────────────────┐ │ CHECK 1: EXISTENCE (mechanical) │ │ Does this cited work exist in paperpile.bib? │ │ → grep paperpile.bib for author + title + year │ │ → If not found: flag as UNVERIFIED (may still exist) │ └──────────────────────────────────────────────────────────┘ │ Exists in bib? ▼ ┌──────────────────────────────────────────────────────────┐ │ CHECK 2: ACCURACY (mechanical) │ │ Are the citation fields correct? │ │ → Compare volume, issue, pages, year against bibtex │ │ → Flag any mismatches as FIELD_ERROR │ └──────────────────────────────────────────────────────────┘ │ Fields correct? ▼ ┌──────────────────────────────────────────────────────────┐ │ CHECK 3: QUOTE VERIFICATION (outsourced to RAG) │ │ Does the quoted text appear in the source? │ │ → readwise chat: "Verify this exact
- What This Skill Checks
- Independence Architecture
- Scope
- Federal Case Verification via WRDS
- Iron Law: Mechanical Checks Before LLM Checks
- Step 0: Prerequisites
- Download paperpile.bib
- Extract Footnotes
- Step 1: Check 1 — Citation Existence
- Matching Strategy
- Classification
- Step 2: Check 2 — Field Accuracy
- Step 3: Check 3 — Quote Verification
- Tier A: Readwise Highlights (fastest — no download needed)
Connect via SSH tunnel to WRDS ssh wrds "echo \"SELECT plaintiff, defendant, docket, district, filedate FROM fjc_litigation.civil WHERE plaintiff ILIKE '%smith%' AND defendant ILIKE '%jones%' AND filedate BETWEEN '2018-01-01' AND '2020-12-31' LIMIT 10;\" | psql -h wrds-pgdata.wharton.upenn.edu -p 9737 -d wrds" Download paperpile.bib from Drive gws drive files get --account eddyhu@gmail.com \ Quick extraction via python-docx pixi exec --spec python-docx -- python3 -c "
What does the source-verify skill do?
This skill should be used when the user asks to 'verify sources', 'check citations', 'fact-check footnotes', 'verify quotes', 'source check', 'are my cites real', 'check for hallucinated citations', or needs to verify that citations in a legal/academic manuscript actually exist and are accurate. Also use when the user wants to check that quoted text actually appears in the cited source.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill source-verify --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.