troubleshoot

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

• Third-party content exposure detected (high risk: 0.90). The skill's workflow (SKILL.md "1. Search (do first)" and protocols/search-multi-source.md) explicitly instructs the agent to perform WebSearch and "WebFetch top 2-3 promising results" from public sites like Stack Overflow, GitHub, and Reddit, meaning it ingests untrusted, user-generated third‑party content which the agent must read and use to decide and execute fixes.