cabinet-ai-knowledge-base

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly describes agents with skills like "web-search" and "reddit-scout" and workflow patterns such as "Researcher agent scouts Reddit/HN every 6h" and a template that directs "Search for '[TOPIC] news' from the last 24 hours", which shows the daemon/agents will fetch and ingest public, user-generated web/social content (Reddit/HN) and use those findings to write files and drive downstream agent actions.