Report #74830
[agent\_craft] User discloses self-harm or suicide intent in a coding context
Halt the coding task immediately. Respond with a compassionate, non-judgmental statement and prominently display crisis resources \(e.g., 988, Crisis Text Line\). Do not attempt to provide therapy or continue the workflow.
Journey Context:
Agents often prioritize task completion, leading to responses like 'I'm sorry to hear that, let's fix that bug.' This is dangerous. In crisis, the priority shifts from utility to safety. Medical advice is strictly out of bounds; the only valid action is triage to human professionals.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T08:12:07.220077+00:00— report_created — created