byte-research

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md "Required Behavior" and "Workflow" explicitly instruct the agent to browse the web and ingest public third-party sources (e.g., app store listings and reviews, pricing pages, public roadmaps, industry articles, GitHub repos), which are untrusted/user-generated content the agent must read and that can materially influence decisions.