continuous-learning

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill runs as a Stop hook that "has access to full session transcript" and automatically extracts and writes patterns to persistent files, so it could capture and then output secrets from sessions verbatim.

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

• Third-party content exposure detected (medium risk: 0.30). [SKILL.md]/evaluate-session.sh uses the Stop hook to process a session transcript transcript_path (read from hook stdin/env) and then extracts patterns from the full session transcript; since an outsider-authored free-text user can enter messages into that session, the workflow ingests untrusted free text at runtime (without needing to select a specific item first, beyond the session being long enough).