bullshit-detector
Fact-check and hype-audit content. Extracts the discrete claims from a video, article, tweet, or PDF, verifies each against independent sources via web search, and produces a report card with per-claim verdicts and an overall BS score (0-10). Use when the user asks to fact-check, verify, debunk, or evaluate credibility — "is this true/legit/bullshit", "check this video", "how much of this holds up".
npx skills add SerhiiKorniienko/bullshit-detector --skill bullshit-detector --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
Fact-check and hype-audit content by extracting each discrete claim from input content (video, article, tweet, or PDF), verifying each claim with independent web sources, and producing a report card with per-claim verdicts and an overall BS score (0-10).
How it works
- Get the text from the input or a fetch-content workflow, preserving metadata (views, author, date), and save a normalized copy to a /tmp path for reuse.
- Read the entire piece before judging, noting the author's incentives.
- Extract claims, listing every distinct assertion and classifying as factual, prediction, opinion, or anecdote, and number them with source timestamps/locations.
- Verify load-bearing claims by performing web searches for independent evidence and ranking findings per a rubric, enforcing that each claim is supported by origin-backed sources. Cap verification as budget allows; mark unchecked claims with ⚪ not checked where applicable.
- For each checked claim, decide verdicts based on evidence; if a gap exists, clearly indicate it in the bottom line.
- Use a structured approach to handle gaps, referents, and ambiguous readings, ensuring each surviving claim stands alone in its own row.
- Write the report card using the rubric template, including a BS score (0-10).
- Save the final report to a file as the artifact.
When to use it
Use when the user asks to fact-check, verify, debunk, or evaluate credibility — such as queries like "is this true/legit/bullshit", "check this video", or "how much of this holds up".
What it can touch
The workflow involves web searches and a verification process, using the defined claims and the rubric for citation and tiering. The exact touchpoints include extracting claims, performing independent web searches, and writing a structured report card.
Caveats
The process emphasizes exhaustive extraction and verification within budget constraints; unchecked load-bearing claims are flagged as not audited. If any evidence cell cites a DOI, a retractions check may be triggered via the specified command in the workflow. The tool uses conditional steps and may cap verification if costs exceed budget.
# bullshit-detector Separate what's verifiably true from what's hype in any piece of content. ## Workflow **Start at step 1 now. The steps below are the plan** — they are already ordered, and each one says what it needs. There is nothing to work out in advance, and working it out anyway is measurably expensive: across 35 instrumented runs the phase before the first tool call is almost entirely del
What does the bullshit-detector skill do?
Fact-check and hype-audit content. Extracts the discrete claims from a video, article, tweet, or PDF, verifies each against independent sources via web search, and produces a report card with per-claim verdicts and an overall BS score (0-10). Use when the user asks to fact-check, verify, debunk, or evaluate credibility — "is this true/legit/bullshit", "check this video", "how much of this holds up".
How do I install it?
Run `npx skills add SerhiiKorniienko/bullshit-detector --skill bullshit-detector --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 SerhiiKorniienko/bullshit-detector, a repository with 104 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.