agent-memory

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

• Third-party content exposure detected (medium risk: 0.30). In the required runtime workflow, the UserPromptSubmit/SessionStart→memory_extract.py path reads outsider-authored free text from Claude Code transcripts by iterating user (human) messages in transcript JSONL and extracting marker-matching spans into L1 atoms that can later be injected into prompts.