fact-claim-audit
Skill2 starsChanged 2 months agoMITView on GitHub
Audit a claim, conclusion, dataset statement, or plan rationale by separating verifiable facts, inference, and value judgment, then checking evidence quality and reasoning. Use for fact checks, disputed or current claims, causal assertions, high-impact evidence, and credibility judgments. Skip pure preferences, fiction, translations, and deterministic questions settled by calculation or tests.
In this repository9 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
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 12
Use when the user asks whether a claim is true, credible, causal, current, or adequately supported.
- Distinctive: line 27
Follow current browsing and citation rules when external verification is required.
- Distinctive: line 34
Every material claim has a status and matching evidence.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Fact and Claim Audit
Own atomic claim verification, evidence quality, reasoning-chain audit, and confidence calibration. Do not replace domain research narratives or decisions.
Route
Use when the user asks whether a claim is true, credible, causal, current, or adequately supported; when a high-impact conclusion depends on external facts; or when a memo, dataset, or plan may mix evidence with inference.
Skip pure preferences, fiction, translation, formatting, and deterministic questions settled by calculation, inspection, or runnable tests. When facts may have changed, precision matters, or risk is high, use current primary sources.
Procedure
- Split the input into minimal decidable claims and label each fact, inference, value judgment, or non-falsifiable wording.
- Define the test for each fact: object, metric, time, location or sample, and required context.
- Find the nearest primary or authoritative record and, when material, corroborate with an independent source. Record publication date and event date separately.
- Rate each claim: verified; basically supported but narrower; disputed; insufficient evidence; or clearly wrong.
- Assuming the supported facts, audit hidden premises, correlation versus causation, alternative explanations, omitted variables, extrapolation range, and failure conditions.
- Rewrite the strongest version the evidence permits, explain confidence, and name the evidence most likely to change the assessment.
Handoffs and Boundaries
- Follow current browsing and citation rules when external verification is required.
$longitudinal-horizontal-analysisowns the full time-and-comparator study; this skill audits its claims.$bidirectional-steelmanowns the contest between conclusions; this skill provides the shared evidence base.- Dataset-wide validation belongs to the applicable data-analysis skill.
Completion Gate
Every material claim has a status and matching evidence. Fact status is separate from reasoning validity, unknown is not treated as false, citations directly support adjacent claims, and confidence has a stated basis.
Source Status
Distilled from the user-supplied “事实核查” method card. The card describes a skeptical workflow but offers no accuracy benchmark or measured performance.
<!-- integration-watermark: 红猴子整合 -->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.