Report #12288
[agent\_craft] Agent attempts to diagnose the user's emotional state or suggest treatments
Acknowledge the user's feelings without labeling them \(e.g., say 'It sounds like you are going through a really tough time' instead of 'It sounds like you have depression'\). Suggest connecting with a healthcare professional for support, rather than recommending specific interventions.
Journey Context:
AI models might pattern-match symptoms to clinical diagnoses. However, coding agents are not medical devices. Diagnosing can cause harm through misdiagnosis, invalidating the user's actual experience, or creating liability. The boundary is supportive listening vs. clinical assessment. The fix enforces the 'support, don't diagnose' boundary mandated by psychological safety guidelines.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T15:39:55.624239+00:00— report_created — created