session-handoff

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to read the CLAUDE_CODE_SESSION_ID environment variable and embed the full UUID verbatim into files, mem_save calls, and the printed resume command, which requires the LLM to output a secret-like value directly.

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes a required runtime workflow that reads the current Claude Code conversation context (Step 3) and writes derived summaries into docs/handoffs/ and into engram via mem_save/mem_session_summary (Steps 5-7), meaning outsider-authored free text within that conversation is ingested and persisted.