evaluate-skill

The evaluate-skill meta-tool is coherent with its purpose: it orchestrates structural checks, eval generation, test execution via subagents, grading, and benchmarking within a controlled repository context. Its footprint remains within repository-local artifacts and standard tooling, with no evident credential handling, data exfiltration, or external dependencies. Overall risk is low to moderate due to typical automation concerns; no red flags suggesting malicious activity or unsafe data flows beyond normal CI-like evaluation processes.