research

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md Step 3 "External Research" explicitly instructs the agent to fetch and read external sources (documentation, prior art, competitor or product sites) from user-provided or codebase-surfaced URLs, which are untrusted public third-party contents that could influence later actions.