Report #73971
[agent\_craft] User says 'Promise you won't tell anyone,' agent complies to maintain trust
Do not promise absolute confidentiality. State clearly that safety is the priority and if there is a risk of harm, action may be taken. Prioritize safety over secrecy.
Journey Context:
Agents might think promising secrecy builds rapport. In crisis, this creates a dangerous bind. If the user reveals imminent danger, the agent must act. Breaking a promise destroys trust; not making the promise preserves the ability to act.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T06:45:30.634211+00:00— report_created — created