In October 2026 we pulled the 12 GitHub repos named in five popular AI videos, cloned each into an isolated folder and read the install hooks, network calls and credential reach. We installed none of them as-is. A “SAFE” verdict means a static read found nothing alarming. It is not a guarantee, and repos change, so read the code yourself before you run anything.
Stars, licences and push dates below were re-read from the GitHub API on 2026-10-15.
Explainers
- no-ai-slop: a writing-edit skill that strips AI tells.
- video-use: lets a coding agent edit real footage.
- SkillSpector: scans agent skills for risky patterns before you install them.
- playwright-mcp: browser control for AI agents.
- agency-agents: a big roster of role prompts.
Get the full list of 12 vetted repos
All 12 repos with verdict, licence, what decided it and what we would use each for. The link comes to your inbox.