gauge

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly performs web research (see SKILL.md's "research" recipe and EVOLVE phase) and references/requires ingesting open public sources (references/web-sources.md lists T3/T4 sources like Reddit, blogs, GitHub, arXiv) whose findings are used to update detection patterns and self-evolution decisions, so untrusted third‑party content can materially influence agent behavior.