what

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow runs what_concepts.py over the current session slice by reading the operator-visible agent prose (and optional tool_use inputs) from the workspace transcript/conversation files on --transcript/--conversation (e.g., via /what invocation), so outsider-authored free text in those logs can be ingested as candidate terms.