fact-checker
Skill3 starsChanged 40 days agoMITView on GitHub
Verifies discrete factual claims against multiple independent sources and returns a calibrated verdict - Verified, Likely true, Likely false, Unverifiable, or False - with confidence and citations. Use when someone asks "is this true", "fact-check this claim", "did X really happen", "verify this statistic before we publish", or pastes a quote, number, or viral post to check. Do NOT use for tracing a citation chain back to its primary origin and appraising study methodology - use claims-verifier instead; for a broad open-ended question that needs a full research brief, use deep-research.
21 other projects use a version of this.
Adopted by
Most recently taken up by
- aditya-vermax/skillstheir own version, first seen Nov 8, 2026
- BrickyStudios/ai-fact-checker-claude-skilltheir own version, first seen Nov 8, 2026
- weininghui/skillstheir own version, first seen Oct 30, 2026
- daymade/claude-code-skillstheir own version, first seen Oct 19, 2026
- yasdelayu/agent-x-cctheir own version, first seen Oct 13, 2026
- IbrahimSobh/AISkillstheir own version, first seen Oct 9, 2026
Used with
Projects with it also use qa-expert, prompt-master and 8 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
belokonmSkillqa-expert
Quality assurance specialist focusing on test strategy, quality processes, and comprehensive testing methodologies
Also in 3 projects that use this
ttmouseSkillprompt-master
提示词主控 - 智能选择合适的领域skill并生成提示词,支持自动领域分类和调度
Also in 3 projects that use this
NVIDIASkilldeep-research
Queue deep, multi-step research and analysis tasks to the DeepAgents worker. Supports execution depth presets, request-specific rubrics, SubAgent delegation, and RubricMiddleware cross-validation.
Also in 4 projects that use this
Microsoft AzureSkillhumanizer
Tone enforcement patterns for external-facing community responses
Also in 4 projects that use this
AnthropicSkillxlsx
Comprehensive spreadsheet creation, editing, and analysis with support for formulas, formatting, data analysis, and visualization. When Claude needs to work with spreadsheets (.xlsx, .xlsm, .csv, .tsv, etc) for: (1) Creating new spreadsheets with formulas and formatting, (2) Reading or analyzing data, (3) Modify existing spreadsheets while preserving formulas, (4) Data analysis and visualization in spreadsheets, or (5) Recalculating formulas
Also in 3 projects that use this
VercelSkillfind-skills
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Also in 3 projects that use this
Other versions21 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 21
- daymade/claude-code-skills★ 1.5ksame name
- IbrahimSobh/AISkills★ 7same name
- xcrrr/claude-skills★ 6same name
- Bartosh16/claude-skills★ 3same name
- BrickyStudios/ai-fact-checker-claude-skill★ 2same name
- ethicalcapital/skills★ 2same name
- seaneoliver/claude-code-starter★ 2same name
- suchwow-sysadmin/claude-code-factchecker★ 2same name
- aditya-vermax/skills★ 1same name
- arielnishri-svg/claude-skills★ 1same name
11 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository5 other files it gives its agents.
- academic-essaySkill
- citation-trackerSkill
- literature-reviewSkill
- policy-briefSkill
- systematic-reviewSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 9 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 19 edited versions we read keep these.
9
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Follow in order - the verdict scale is meaningless if the claim was never isolated.
- Distinctive: line 26
Gather sources: Prefer primary sources - official data, original papers, direct statements - over summaries of them.
- Distinctive: line 29
Assign verdict and confidence: Use the scale below, state confidence (high / moderate / low), and state what specific evidence would change …
- Distinctive: line 35
Unverifiable: - no adequate sourcing exists.
- Distinctive: line 55
Do not fabricate or embellish a citation - a wrong verdict is recoverable, an invented source destroys all trust in every verdict.
- Distinctive: line 56
Do not count repetitions as corroboration.
- Distinctive: line 57
Do not grade plausibility.
- Distinctive: line 58
Do not call an unverified claim False.
- Distinctive: line 59
Do not check the spin. Verify the factual core and say explicitly which interpretive framing was set aside.
“Common” is measured: kept by at least 6 in 10 of the 19 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Fact Checker
Treat every claim as unproven until corroborated. The job is calibrated truth, not a verdict for its own sake - and the costly failure this skill prevents is the confident verdict built on correlated sources: three outlets repeating one wire story feels like consensus and is actually a single point of failure. "I can't verify this" is a valid, honest answer; a fabricated citation is never acceptable.
Inputs to collect
Gather these before checking anything. If the user cannot supply one, use the default and label the assumption.
- The exact claim, verbatim. Paraphrase later; anchor on the original wording first.
- Where it appeared and what is at stake (publishing it, acting on it, arguing against it). Stakes set the sourcing bar - anything being published gets the full two-independent-primary-sources standard.
- Depth and deadline. Default: full check.
- Any sources the user already has. Treat these as leads, not evidence, until appraised.
Operating procedure
Follow in order - the verdict scale is meaningless if the claim was never isolated.
- Isolate the claim. Restate it as one or more precise, checkable propositions. Split vague or compound claims into parts and give each part its own verdict. Separate the factual core from interpretation or spin; check the core.
- Gather sources. Prefer primary sources - official data, original papers, direct statements - over summaries of them. Find at least two independent ones per proposition before any strong verdict.
- Check independence. Three outlets citing the same wire story is one source, not three. Same press release, same underlying dataset, same original interview: one source. A company or person making a claim about itself is one interested source, whatever its reach.
- Weigh evidence. Note recency, authority, method, and conflicts of interest. Weigh sources; never average them - one rigorous primary source outweighs five repetitions of a rumor.
- Assign verdict and confidence. Use the scale below, state confidence (high / moderate / low), and state what specific evidence would change the verdict.
Verdict scale and thresholds
- Verified - at least two strong, independent, primary sources agree and no credible source contradicts. Never issue Verified on a single source, however good.
- Likely true / Likely false - leaning verdict; evidence is thin, mixed, or single-source. A single-source claim is capped here.
- Unverifiable - no adequate sourcing exists; say so plainly rather than guessing.
- False - credible independent sources directly contradict it. Same two-source bar as Verified.
Flag any statistic whose underlying data is more than a few years old or has been superseded; a true-then claim can be a false-now claim.
Worked example
Claim as submitted: "The Great Wall of China is visible from space with the naked eye."
- Isolated propositions: (a) visible unaided from low Earth orbit; (b) visible from the Moon (the common stronger version).
- Sources: NASA Earth Observatory analysis of astronaut photography (primary - agency analysis of original imagery); direct statement from astronaut Yang Liwei, China's first person in space, that he could not see it (primary - firsthand observation); Apollo crew accounts that no human structure is distinguishable from the Moon (primary, independent of the first two).
- Independence check: different agencies, different individuals, different underlying observations - three genuinely independent sources.
- Verdict: False for both propositions, high confidence. What would change it: a documented, verified unaided-eye sighting from orbit under stated conditions.
Deliverable
Produce a verification report containing: the claim as restated (with parts, if split); the source list, each with enough detail to locate it and a one-line independence/authority note; the verdict per proposition; confidence with reasoning; and a "what would change this verdict" line.
Do NOT
- Do not fabricate or embellish a citation - a wrong verdict is recoverable, an invented source destroys all trust in every verdict.
- Do not count repetitions as corroboration; trace what each outlet actually relied on.
- Do not grade plausibility. A claim that sounds right and cannot be sourced is Unverifiable, not Likely true.
- Do not call an unverified claim False. Absence of evidence is not contradiction.
- Do not check the spin. Verify the factual core and say explicitly which interpretive framing was set aside.
Quality bar
Before shipping, confirm: every Verified or False verdict rests on at least two independent sources named in the report; every source is findable from the citation given; confidence is stated with the condition that would change it; compound claims were split and each part carries its own verdict; the factual core was separated from interpretation.
Escalation
A verdict on facts is not legal, medical, or financial advice - flag claims in those domains for professional review before anyone acts on them. When the real job is following a citation chain to its origin and appraising study methodology, hand off to claims-verifier. When the question is a whole topic rather than a discrete claim, use deep-research; for academic literature, literature-review.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

