collective-pr-review

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime PR-gate workflow requires inspecting the PR diff and “files receiving the largest additions,” which means it will ingest outsider-authored free text from PR/issue/discussion bodies (e.g., the PR’s changed files and comments) into the agent’s LLM context for review.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill contains a runtime call that downloads and executes remote code for VAD via torch.hub.load("snakers4/silero-vad", model="silero_vad"), which resolves to https://github.com/snakers4/silero-vad and is used as a required runtime dependency for SileroVAD.