review-pr

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md ingests outsider-authored free text from external PRs by running gh pr view ... --json title,body,... and then uses gh pr diff ... to scope and launch reviewer/tester skills on the changed code.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches PR metadata and diffs at runtime for PRs identified by the supplied URLs (e.g., https://github.com/{owner}/{repo}/pull/{number} and https://gitlab.com/{owner}/{repo}/-/merge_requests/{number}), and those fetched diffs are injected into agent skill prompts as the verification scope, so external content directly controls agent prompts.