debug-mode

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

• Third-party content exposure detected (medium risk: 0.65). The workflow ingests outsider-authored free text only if users select checkpoint option C (checkpoint templates in references/checkpoints.md), which is then treated as checkpoint evidence and routed by the agent into the LLM context; this is free-form user-entered text on an interaction boundary rather than trusted logs.