# AI use cases for Bullshit Detector (SerhiiKorniienko/bullshit-detector)

Repo: [SerhiiKorniienko/bullshit-detector](https://github.com/SerhiiKorniienko/bullshit-detector) | Use cases: 4 | Page: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/

A set of portable Agent Skills that fact-check YouTube videos, TikTok videos, articles, tweets, PDFs, and pasted drafts claim by claim, with sources and a 0-10 BS score.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Fact-check a viral video claim by claim before you act on it: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/fact-check-viral-video-before-sharing/
- Fact-check your own post or README before you publish it: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/fact-check-own-draft-before-publishing/
- Summarize a long video or explain one timestamp instead of watching it: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/summarize-long-video-or-explain-one-part/
- Install the skills as a Claude Code plugin that stays current: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/install-as-claude-code-plugin-auto-updates/

## How to fact-check a viral video claim by claim before you act on it with Bullshit Detector

Repo: [SerhiiKorniienko/bullshit-detector](https://github.com/SerhiiKorniienko/bullshit-detector) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Get per-claim verdicts, sources, and a 0-10 BS score for a YouTube or TikTok video.

**Who it is for:** A person who is about to act on, buy into, or share a viral 'make money with AI' or finance video and wants the claims checked first.

**The problem:** Views, production value, and confidence are not evidence, and checking each claim by hand takes far longer than watching the video.

**The outcome:** A report that lists each claim with a verdict (confirmed, plausible, misleading, false, or unverifiable), a hype-signal scan, an incentive analysis, and a 0-10 BS score.

### Steps

1. Install uv from https://docs.astral.sh/uv/ if you don't have it.
2. `npx skills@latest add SerhiiKorniienko/bullshit-detector`
3. Open your agent and ask: is this bullshit? <video URL>
4. Read the BS score and the per-claim verdicts.
5. Open the sources linked for any claim you plan to act on and confirm them yourself.

### Prompt to paste into your agent

```text
Using the bullshit-detector skills in SerhiiKorniienko/bullshit-detector, fact-check the video at <video URL> claim by claim. Return each claim with its verdict, its sources, and the overall 0-10 BS score. Verify that every verdict links to a source rather than relying on model memory.
```

### Caveat

Web search is non-deterministic, so verdicts can change between runs and a single report is one reading, not a measurement.

Risk, remote script: The skills.sh installer copies skills from this GitHub repo, and the fetch script uses uv to resolve its own dependencies, so review the repo before running it on a machine you care about.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/fact-check-viral-video-before-sharing/

## How to fact-check your own post or README before you publish it with Bullshit Detector

Repo: [SerhiiKorniienko/bullshit-detector](https://github.com/SerhiiKorniienko/bullshit-detector) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Flag the claims a hostile reader would challenge first, with a source that fixes each one.

**Who it is for:** An author, founder, or marketer about to ship a post, README, or launch announcement who wants to catch factual errors before readers do.

**The problem:** A stale price, an overstated setup time, or an unsourced statistic in a draft is cheaper to fix before publishing than after a correction.

**The outcome:** A list of the draft's risky claims, each with the source that supports or corrects it, so you can edit before you post.

### Steps

1. Install uv from https://docs.astral.sh/uv/ if you don't have it.
2. `npx skills@latest add SerhiiKorniienko/bullshit-detector`
3. Open your agent and paste your draft, asking: fact-check my draft.
4. Review each flagged claim and the source attached to it.
5. Edit the draft to match the sources, then publish.

### Prompt to paste into your agent

```text
Using the bullshit-detector skill in SerhiiKorniienko/bullshit-detector, fact-check the draft text I paste below. List the claims a hostile reader would challenge first, each with the source that supports or corrects it. Verify that no claim is marked confirmed without a source.
```

### Caveat

The checker can only cite sources it can reach, so a clean report does not prove a draft is accurate.

Risk, remote script: The skills.sh installer copies skills from this GitHub repo, so review the repo before running it on a machine you care about.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/fact-check-own-draft-before-publishing/

## How to summarize a long video or explain one timestamp instead of watching it with Bullshit Detector

Repo: [SerhiiKorniienko/bullshit-detector](https://github.com/SerhiiKorniienko/bullshit-detector) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Turn a YouTube, TikTok, article, or PDF link into a summary or a clear explanation of one part.

**Who it is for:** A busy reader or viewer who needs the gist of a 27-minute video or long article, or the meaning of one confusing moment, without watching the whole thing.

**The problem:** Agents cannot watch videos, and YouTube's official API will not give captions for videos you do not own, so the content has to be fetched as text first.

**The outcome:** A plain-language summary of the link, or an explanation of the part you name, built from the fetched transcript or text.

### Steps

1. Install uv from https://docs.astral.sh/uv/ if you don't have it.
2. `npx skills@latest add SerhiiKorniienko/bullshit-detector`
3. Open your agent and ask: summarize <URL>
4. Or ask: explain the part at 12:30 in <URL>
5. If a site blocks fetching, paste the text or transcript and ask again.

### Prompt to paste into your agent

```text
Using the fetch-content skill in SerhiiKorniienko/bullshit-detector, fetch <URL> and summarize it in plain language. If I name a timestamp, explain that part. Verify that the summary reflects the fetched text and says so if captions or text were unavailable.
```

### Caveat

Fetching depends on what each site exposes, and the tool reports an actionable hint such as paste the text when a paywall or missing captions blocks it.

Risk, remote script: The skills.sh installer copies skills from this GitHub repo, so review the repo before running it on a machine you care about.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/summarize-long-video-or-explain-one-part/

## How to install the skills as a Claude Code plugin that stays current with Bullshit Detector

Repo: [SerhiiKorniienko/bullshit-detector](https://github.com/SerhiiKorniienko/bullshit-detector) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README

Use a managed, read-only bundle that updates as new versions ship.

**Who it is for:** A Claude Code user who wants the fact-checking skills to work without maintaining copied files, and who will not also install the skills.sh copy.

**The problem:** Copied skills drift from the upstream repo over time, and keeping them current by hand is extra work.

**The outcome:** A Claude Code plugin bundle of the skills that updates when a new version ships, ready for fact-checking prompts.

### Steps

1. Open Claude Code.
2. `/plugin marketplace add SerhiiKorniienko/bullshit-detector`
3. `/plugin install bullshit-detector@serhii-korniienko`
4. Ask: is this bullshit? <URL>
5. Do not also run the skills.sh installer, because two copies give Claude Code two copies of every skill.

### Prompt to paste into your agent

```text
In Claude Code, add the marketplace SerhiiKorniienko/bullshit-detector and install bullshit-detector@serhii-korniienko. Then fact-check <URL> and return the per-claim verdicts and the 0-10 BS score. Verify that only one copy of each skill is installed.
```

### Caveat

The plugin is read-only and always current, so you cannot edit the skills the way you can with the copied skills.sh install.

Risk, remote script: The plugin pulls the skills from this GitHub marketplace, so review the repo before installing it.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/serhiikorniienko-bullshit-detector/install-as-claude-code-plugin-auto-updates/