skills-manager

The code documentation describes a legitimate multi-agent skill manager with expected filesystem and execution behaviors. There is no explicit malicious code in the provided text. However, recommended workflows (npx downloads, direct GitHub installs, global installs, and -y automated installs) create a moderate supply-chain risk because they enable execution of untrusted code with broad filesystem and runtime privileges without built-in verification or sandboxing. Treat installs from remote sources as untrusted: add signature/checksum verification, avoid automatic global installs, and require manual review or a sandboxed validation step before executing skill-provided scripts.