prowl-cli

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

• Third-party content exposure detected (high risk: 0.75). Runtime LLM context exposure is via prowl read --json / prowl send --capture --json, which ingest rendered terminal text from the target pane (i.e., free-form output authored by other processes/users), and that text is then available to the agent/LLM through the CLI’s JSON fields like .data.text / .data.capture.text.