A retrieval-gated skill architecture for LLM agents that scales to hundreds of tools by exposing only the top-K relevant capabilities per request.
SaferSkills independently audited SkillMesh (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Route this request with SkillMesh: <your task>Use scripts/route.sh (or scripts/route.py) to emit context:
scripts/route.sh --provider codex --registry <path> --query "<query>" --top-k 5List available roles and install a role bundle (role card + missing dependency cards):
skillmesh roles
skillmesh roles list
skillmesh Data-Analyst install
skillmesh roles install Data-AnalystUse --dry-run to preview install changes.
If skillmesh is not on PATH, use: scripts/roles.sh (interactive)
--registry: Registry file path (tools.yaml, tools.json, or roles.registry.yaml).--provider: codex (default) or claude.--top-k: Number of cards to retrieve (default 5).--backend: auto, memory, or chroma (default auto).--dense: Enable optional dense reranking.--instruction-chars: Max instruction chars per expert (default 700).--query: User request text.--registry is omitted, the router uses SKILLMESH_REGISTRY.skillmesh is not on PATH, the router falls back to python -m skill_registry_rag.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.