signal-review

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

• Third-party content exposure detected (high risk: 1.00). The skill's Scope Inference explicitly instructs the agent to fetch and review PR content using "gh pr diff {url}" for a provided PR URL, which causes the agent to ingest and act on arbitrary third-party (user-generated) web content.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly states that if given a PR URL it will "fetch with gh pr diff {url}" at runtime (e.g. a GitHub PR like https://github.com/owner/repo/pull/123), which injects externally fetched diff content into the agent's input and thus directly controls prompts — this meets the conditions for a runtime external dependency risk.