code-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 workflow ingests the PR/MR diff and file surface (via the “Compute the diff once from the selected base branch” Tier 0 step), which is outsider-authored code/text from other contributors, and that diff-derived prose is then fed into the agent’s LLM context for review.