skill-idea-miner

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text from the user’s external messages in Claude Code session logs is read at runtime from ~/.claude/projects/<project>/*.jsonl (via parse_session() → user_messages) and then injected into the LLM prompt in abstract_with_llm() (_build_llm_prompt() includes ## User Message Samples).