A new tool, bullshit-detector, offers claim-by-claim verification for online content, assigning a 0-10 BS score to YouTube videos, articles, tweets, and PDFs. Developed by Serhii Korniienko, the agent skills are built with Claude Code and utilize plain markdown and self-contained Python.

Installation is available via a `skills.sh` script, which copies skills for customization, or as a read-only Claude Code plugin for a managed, up-to-date experience. The system separates content fetching from analysis, allowing for consistent verification across diverse sources like YouTube, TikTok, and articles, even when direct API access is limited.

The BS score and accompanying report aim to combat misinformation by extracting claims, verifying them against independent sources, and identifying potential biases. Examples show its application to a "make money with AI" video (5/10 BS score) and a TikTok video about the sun (9/10 BS score), demonstrating its ability to dissect hype and fabricated claims.