session-log-analyzer

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md Step 1/Step 2 has the workflow read local Claude Code session logs from ~/.claude/projects/{encoded-path}/*.jsonl and parse their user/assistant/system/progress JSONL contents, which include outsider-authored messages (i.e., the user’s inputs) at runtime.