agent-skill-observatorylisted

# Agent Skill Observatory Evaluate, audit, score, and discover AI Agent Skills using strict static analysis and evidence-based metrics without executing untrusted code. ## Invariants 1. **Static Analysis Only**: Never execute third-party scripts, binary binaries, or curl-pipe-bash installers during indexing or analysis. 2. **Provenance Preservation**: Every indexed skill points back to its repository URL, branch, path, and discovery source. 3. **Multi-Dimensional Scoring**: Quality, Security, Maintenance, and Adoption scores remain separate and explainable. Never collapse into an unexplained vanity metric. 4. **Manifest Grounding**: A candidate repository requires a verified `SKILL.md` manifest before being recognized as a skill. --- ## Capabilities & Workflows ### 1. Scan Local Skill Directory Before Installing Audit a candidate skill directory for structural validity, resource inventory, and static security risks: ```bash skillobs scan-local /path/to/candidate-skill ``` Expected output: ```json { "name": "candidate-skill", "spec": { "valid": true, "errors": [], "warnings": [] }, "security": { "score": 100, "findings": [] }, "score": { "overall": 82, "quality": 85, "security": 100, "maintenance": 90, "adoption": 45 }, "resources": { "scripts": 1, "references": 2, "assets": 0, "other": 0 } } ``` ### 2. Inspect Catalog Evidence Query the verified index via the local or remote API: ```bash # Query skills matching a capability or keyword