670,000 agent skills, no trust layer: bot scan finds 69% never reliably fire
markjeffrey · x · 2026-09-22
DavFields pointed a bot at 14 sources (HN, GitHub, Reddit, X, YouTube) collecting 2,719 posts in 48 hours about agent skills: 960 were feature requests, 526 complaints about broken installs. Key findings: a static scan of 216 public Claude Code skills found 69% don't reliably trigger; security scanners contradict each other on the same skill (Gen: Safe, Socket: 0 alerts, Snyk: Critical Risk), and Snyk's own audit flags 36.8% of 3,984 skills with at least one issue. Conclusion: contradictory trust signals are worse than none across 670,000 skills with no trust layer.
More from coding & agent
- A coding agent prompt: use my dotfile, keep keys safe, test locally without help — yoobinray · 2026-09-22
- Why devs skip Tinker for posttraining: cost and owning your own code — silver__tsuki · 2026-09-22
- codebase-memory-mcp indexes codebases into a persistent knowledge graph in milliseconds — tom_doerr · 2026-09-22
- Ambion: open-source TypeScript collaboration kernel for agents and humans, formally verified with Dafny — andreisavu · 2026-09-22
- Ambion: A Collaboration Kernel Running Multiple Agent SDKs in One Shared Room — andreisavu · 2026-09-22
- Formal Verification and Chaos Testing for Multi-Agent Systems with Dafny — andreisavu · 2026-09-22