harness-master

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

• Third-party content exposure detected (high risk: 0.80). The skill's required workflow (Gate 2: "Refresh latest official guidance" in SKILL.md) and its references/latest-doc-sources.md explicitly instruct the agent to fetch and read public vendor docs and web fallback URLs (e.g., developers.openai.com, code.claude.com, cursor.com) and to use those sources to drive audit findings and apply decisions, which exposes the agent to untrusted third-party content that can materially influence actions.