anvil

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The run.sh script explicitly fetches and executes a package from a public GitHub URL (exec uvx --from "git+https://github.com/grahama1970/anvil.git" anvil), so untrusted third‑party repository content would be ingested and could change runtime behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The run.sh contains exec uvx --from "git+https://github.com/grahama1970/anvil.git" which, at runtime, fetches and executes code from that remote GitHub repository (git+https://github.com/grahama1970/anvil.git), so the skill depends on and runs external code.