Discover

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

• Third-party content exposure detected (high risk: 0.90). SKILL.md's "External Endpoints" explicitly names public sources (https://www.google.com, https://news.google.com, https://www.reddit.com, and "topic-specific public primary sources") and the discovery/heartbeat workflows (discovery-loop.md and heartbeat-rules.md) require the agent to search and ingest those public web sources to produce findings and next-step recommendations, which exposes the agent to untrusted, user-generated third‑party content that could carry indirect prompt injections.