Skip to content
SaferSkills Docs

About

SaferSkills is a public, free, Apache-2.0 service that independently scans every AI capability — skills, MCP servers, hooks, plugins, and rules — across every agent platform. Anyone submits a GitHub URL or uploads files, a deterministic scan runs in about thirty seconds, and the result is a public report: a 0–100 trust score, a full rule trace, a quotable line of evidence per finding, a vendor right-of-reply, and a permanent permalink. No LLM sits in the verdict path.

What does SaferSkills actually do?

Section titled “What does SaferSkills actually do?”

It scans AI capabilities and publishes the result as a transparent, reproducible record. The engine walks a repository, discovers each capability, and scores it independently against a documented rule set across five sub-scores — Security, Supply Chain, Maintenance, Transparency, and Community. Beyond these static component scans, SaferSkills also runs a behavioral Agent Scan that grades how a whole running agent behaves against a test pack. Everything is deterministic: the same input always produces the same score, and a vendor can re-derive any historical verdict offline.

What does SaferSkills stand for?

Section titled “What does SaferSkills stand for?”

Three principles shape every page and every verdict. Methodology over opinion — each rule is published, each score shows its math, and “we score X by rule Y” replaces “X is bad.” Anti-recommendation — SaferSkills never tells you which capability to use; a low score means review before use, not don’t use. Transparency over erasure — findings are appealable through a vendor right-of-reply but never silently deleted, so the public record stays honest.

Who stewards SaferSkills?

Section titled “Who stewards SaferSkills?”

SaferSkills is an OpenLatch project, run brand-independently. The code is open source under the Apache License 2.0, the methodology is contributed and disputed in public, and the live service runs at saferskills.ai. The source, the rule corpus, and the contribution and appeal templates all live in the public repository at github.com/OpenLatch/saferskills.

Where do I go next?

Section titled “Where do I go next?”