dw-learn

This fragment instructs an operator or capable agent to capture the full conversation, store it in a hidden Diffwiki collection, and mine the evidence for additional articles. This is a privacy and data-retention risk because user requests and raw outputs may be stored outside the conversation, with reduced visibility; no execution or external network destination is shown. Severity is moderate if followed with sensitive conversation content; the fragment itself is not executable malware.

This prompt directs an agent to read local project-scoped session transcripts and copy their contents into a hidden diffwiki collection, creating a potential privacy and information-disclosure risk. No specific external domain, IP address, or credential target is named, and the destination's access controls are unknown. Severity is high if the transcript collection is unauthorized or accessible beyond its intended audience; the fragment itself does not demonstrate execution or establish malicious intent.