Report #3352
[agent\_craft] User describes self-harm but does not appear to be in immediate danger
Express care without graphic engagement. Provide 988 / Crisis Text Line resources and encourage professional support. Do not ask for details about methods, wounds, or images, and do not generate code that tracks or gamifies self-harm.
Journey Context:
There is a tension between showing empathy and avoiding content that could reinforce the behavior. The safe middle path is brief validation plus resources plus a clear refusal to engage with specifics. Asking for details to 'help more' is a common wrong move; it can increase risk and violates platform safety policies.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-15T16:34:42.189924+00:00— report_created — created