claude-code-review

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md, the runtime path takes user-provided review targets (e.g., a PR number/URL) in Step 1/3 and then requires Claude Code CLI to fetch/read the resulting code diff/changed files in Step 4 and during the debate loop, meaning outsider-authored PR text and file contents are ingested at runtime.