cowork-mem

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

• Third-party content exposure detected (high risk: 0.70). The skill's PostToolUse hook (scripts/post_tool_capture.py) explicitly captures Bash command outputs — including commands like curl/wget — and stores them in the local memory DB which is then read by session-start, search, and vector_search to influence decisions, so arbitrary public web content fetched via such tools can be ingested and indirectly steer agent behavior.