gauge

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly runs a web RESEARCH/EVOLVE phase (see "EVOLVE" / "RESEARCH → EVALUATE → CLASSIFY → UPDATE → VERIFY → PERSIST" in SKILL.md) and references/recommends scraping public sources in references/web-sources.md (including T3 community sources like reddit.com/r/ClaudeAI and skillsmp.com), and those web-sourced findings are used to update detection patterns and checklist rules—meaning untrusted third-party content is ingested and can materially change agent behavior.