Fact Check
Verify a user claim by decomposing it, searching web and Semantic Scholar, reading primary sources, and returning cited verdicts per sub-claim.
Overview
Fact-check is an agent skill most often used in Idea (also Validate, Launch) that verifies claims via decomposed sub-claims, web and academic search, and cited Supported/Unsupported verdicts.
Install
npx skills add https://github.com/collaborative-deep-research/agent-papers-cli --skill fact-checkWhat is this skill?
- Five-step workflow: Decompose → Search → Verify Sources → Assess → Report
- paper-search against Google web, Semantic Scholar snippets, and paper lists (--limit 5)
- paper read and paper-search browse for primary-source verification
- Four verdict buckets: Supported, Partially supported, Unsupported, Uncertain
- Mandatory citation rule: no verdict without specific sources and URLs
- 5-step fact-check workflow (Decompose through Report)
- 4 verdict categories per sub-claim
- Semantic Scholar paper search default --limit 5
Adoption & trust: 1 installs on skills.sh; 44 GitHub stars; 2/3 security scanners passed (skills.sh audits); trending (+100% hot-view momentum).
What problem does it solve?
You need to publish or decide on a specific claim but do not know whether evidence from reliable primary sources actually supports it.
Who is it for?
Solo builders validating technical, scientific, or market statements before shipping content, specs, or investor narratives.
Skip if: Situations requiring certified legal review, classified sources, or instant breaking-news confirmation without search tooling.
When should I use this skill?
The user asks to verify, fact-check, or confirm a statement (argument-hint: claim to verify).
What do I get? / Deliverables
You receive a structured report with per sub-claim verdicts, an overall assessment, and URLs to the sources your agent inspected.
- Sub-claim list with Supported/Partially supported/Unsupported/Uncertain verdicts
- Overall assessment with cited URLs
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Fact-checking is canonical on Idea/research where you validate market, science, and competitor assertions before committing—but the same workflow supports content and docs later. Research subphase is where decomposed sub-claims and evidence gathering belong before scope and build decisions.
Where it fits
Decompose a competitor's uptime claim and search papers plus web sources before you cite it in a comparison page.
Verify whether a proposed feature depends on published benchmarks before locking MVP scope.
Fact-check statistics in a long-form article so AI-search snippets stay defensible.
How it compares
Structured evidence workflow with academic CLI search—not a generic web-scrape skill or single-shot LLM 'true/false' answer.
Common Questions / FAQ
Who is fact-check for?
Indie founders, technical writers, and agent users who need cited verification of specific statements using web and Semantic Scholar sources.
When should I use fact-check?
In Idea/research before trusting a stat; in Validate when testing positioning; in Launch/Grow when confirming claims in SEO copy or release notes—the trigger is 'verify, fact-check, or confirm' a statement.
Is fact-check safe to install?
It uses Bash and network search tools—check the Security Audits panel on this page and restrict credentials before granting shell on production machines.
SKILL.md
READMESKILL.md - Fact Check
Fact-check the following claim: "$ARGUMENTS" ## 1. Decompose Break the claim into specific, verifiable sub-claims. List them explicitly before searching. ## 2. Search for Evidence For each sub-claim: ``` paper-search google web "<sub-claim as question>" paper-search semanticscholar snippets "<sub-claim keywords>" paper-search semanticscholar papers "<sub-claim keywords>" --limit 5 ``` ## 3. Verify Sources For each promising source: ``` paper read <arxiv_id> <relevant section> # for academic papers paper-search browse <url> # for web pages ``` Prefer primary sources (original papers, official data) over secondary reports. ## 4. Assess For each sub-claim, assign a verdict: - **Supported**: strong evidence from multiple reliable sources - **Partially supported**: some evidence, with caveats - **Unsupported**: no evidence found, or evidence contradicts the claim - **Uncertain**: insufficient evidence to judge ## 5. Report Present: - The original claim - Each sub-claim with its verdict and supporting evidence - An overall assessment - All sources cited with URLs ## Guidelines - Always cite specific sources — never state a verdict without evidence. - Distinguish between "no evidence found" and "evidence contradicts." - Note the quality and recency of sources. - If a claim is about a specific paper, read that paper directly rather than relying on summaries.