assess-pull-request

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

• Third-party content exposure detected (high risk: 1.00). This skill explicitly instructs the agent to fetch and read provider-backed PR/MR content (e.g., "Use read-only provider commands for evidence when available: gh for GitHub`" and the "Provider-backed mode" section that requires PR/MR pages, diffs, commits, and comments), which are untrusted, user-generated third-party sources that the agent must interpret and which can materially influence its review decisions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly requires fetching repository files at runtime (via provider APIs or git clone), e.g. the example PR URL https://github.com/example/acme-widget/pull/42 and git clone patterns like https://github.com//.git, and those fetched files (plan-task/SKILL.md, review-guidance.md) directly control the agent's review prompts/formatting.