log-to-dosu-knowledge

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow reads and digests local Cursor/Claude/Codex JSONL agent logs from paths under ~/.cursor/projects/.../agent-transcripts, ~/.claude/.../projects/.../*.jsonl, and ~/.codex/sessions/.../*.jsonl (via parse_agent_logs.py), which can contain outsider-authored free text submitted in those agents’ user turns.