reviewing-code

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

• Third-party content exposure detected (high risk: 0.75). The skill’s workflow ingests “diff, issue, PR/MR description, tests, and surrounding code” at runtime; those artifacts are commonly authored by outsiders (e.g., PR/issue text from other users) and are read as free-form prose into the agent’s LLM context.