defragance

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

• Third-party content exposure detected (high risk: 0.85). The skill’s workflow requires selecting “3-5 high-fit comparable repositories or artifacts” and then distilling their “naming, boundaries, entrypoints, file layout, and dependency direction,” which at runtime implies the agent reads outsider-authored repository/docs text (e.g., GitHub/public repos) into the LLM context via fetched repository content or API-delivered issue/wiki/discussion/doc text.