Fact-check a project's own README before publishing claims
At a glance
A skill author ran their claim-checking agent against its own README and corrected misleading statements exposed by the report. This workflow is listed under Documentation, Fact Checking, Agent Skills.
- Agents / tools
- Bullshit Detector, AI coding agent, uv, skills.sh, Web search
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Use an agent that can run scripts and search the web, with Node's npx and the uv package runner available. 2) Install the author's bundle with `npx skills@latest add SerhiiKorniienko/bullshit-detector`, choosing the skills and agent in the installer. 3) Give the agent your README URL or draft text and request a fact-check. 4) Let it retrieve the content, separate factual claims, and search for supporting or conflicting evidence for each claim. 5) Read the resulting sourced report and investigate the misleading or unverifiable items. Correct the README statements supported by that review before publishing.
How it went
In the HN discussion, the author reported correcting two misleading README lines. The repository describes stale API pricing and an understated setup prerequisite among the self-audit findings.
Worth knowing
The author says the tool checks factual premises, cannot reliably judge source truthfulness, and can miss faulty reasoning built on true facts.
