Bullshit Detector skill
Bullshit Detector checks claims from a video, article, tweet or PDF against sources and returns a report scored 0 to 10.
- 163 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skills run a Python script and fetch pages from the network.
- Links and texts of checked materials go to web search.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/bullshit-detectorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills run a Python script and fetch pages from the network.
- Links and texts of checked materials go to web search.
This is third-party code. Review the repository files before installing.
What it does
The skill set breaks a link to a viral YouTube video, article, tweet or PDF into separate claims and checks each by search. The report gives a verdict per claim: confirmed, plausible, misleading, false or unverifiable. With no source there is no verdict, even when the model is sure. An overall score from 0 to 10 tops the report. The skills are portable: Markdown and self-contained Python in Claude Code, Codex, GitHub Copilot, Cursor, Gemini CLI, OpenCode and Zed. Web search and `uv` for the fetch script are required. You can ask for a short summary or a breakdown of the part at 12:30, get an HTML report card and social posts.
Who it is for. Editors, researchers and anyone checking viral videos and loud AI articles before retelling them.
Good fit when
- You must check a video like 10 ways to make money with AI before reposting.
- An article or PDF contains dozens of figures to verify against sources.
- You need an HTML report card to send to a colleague.
- You need a short summary of a video and an explanation of the part at 12:30.
Not a fit when
- The agent has no web search: checking relies on sources found.
- You need a legal or medical verdict: the report checks claims and does not replace an expert.
- A private document is being checked: pages and links go to search.
Example request
Fact-check this video and return a report on each claim with a score.Limitations
The skills read pages and run a self-contained Python fetch script, so `uv` and network access are needed. For TikTok and YouTube without a terminal, Claude Cowork needs the Chrome connector or a pasted transcript. Some features such as posts and the carousel need your machine. Do not install the plugin and copies of the skills together: Claude Code would get two instances. The check relies on sources reachable by search: closed, paywalled and very fresh materials may stay without a verdict. A model computes the 0 to 10 score by the skill rules, so treat it as a guide and open the links in the report.
How to disable. Delete the copied skills from the agent skills directory or run `/plugin uninstall bullshit-detector@serhii-korniienko`.
FAQ
What verdicts exist?
Confirmed, plausible, misleading, false and unverifiable. Without a source no verdict is given.
How do I install it as a Claude Code plugin?
With `/plugin marketplace add SerhiiKorniienko/bullshit-detector` and `/plugin install bullshit-detector@serhii-korniienko`. The plugin updates itself.
How does the skills.sh install differ?
It copies the skills into your setup so you can edit them. The plugin stays a read-only bundle. Pick one way.
Does it work without a terminal?
Yes, Claude Cowork installs the bundle from a GitHub URL; in Claude chat you can paste text or attach a file.
Is there a sample report?
Yes, the repository has a report on an AI bubble video: sixty claims, forty-four checked by individual searches.
Discussion
No questions or reviews yet.
Related
SkillsA skill that audits accounts and viral posts on Xiaohongshu, Douyin, Kuaishou, WeChat Channels and Bilibili, then drafts a copy.
Three experimental skills: adapt an article for channels with a source map, sort exported AI answers for GEO and spot-check hreflang
14 skills and a CLI save videos, podcasts and articles to a local Markdown base and find quotes in it with a source link
A skill that checks a video for Douyin, Xiaohongshu and Bilibili before publishing and returns a five-level violation report.
A skill for Claude Code, Codex and Cursor: strips signs of AI writing and checks the result against human-writing ranges.
A skill that analyses the video behind a Bilibili @ mention and posts a reply with clickable timestamps from your account.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Bullshit Detector (this card) | no | 163 | |
| my-skills by akitaonrails | no | 302 | a personal collection of twenty-four skills where fact-checking is one of many; here a dedicated set with verdicts and sources |
| facts | no | 207 | facts keeps a file of verifiable claims about a project; Bullshit Detector checks external videos and articles |
README in short
The repository explains a set of fact-checking skills: a quick start through skills.sh, plugin install, a feature table by environment and a verdict description. Sample reports and a release subscription link are included. MIT licensed.