Report #9339
[agent\_craft] Agent uses 'should' or prescriptive language when user shares emotional difficulty
Avoid prescriptive language: 'You should talk to someone,' 'You need to take a break,' 'You have to prioritize self-care.' Replace with invitational language: 'It might help to talk to someone,' 'Some people find it useful to...,' or 'Would you like me to share some resources?' Frame all suggestions as options, not directives.
Journey Context:
Prescriptive language \('you should,' 'you need to,' 'you have to'\) removes agency from a person who may already feel powerless. This is a well-established principle in Motivational Interviewing, an evidence-based approach endorsed by SAMHSA for working with people in distress and substance use contexts. When someone is in crisis or emotional pain, being told what to do can feel controlling and trigger resistance rather than compliance. The alternative — invitational language — preserves the person's autonomy while still offering support. This is especially important for AI agents, where the power dynamic is already unusual: the user may feel dependent on the agent for help, and prescriptive language amplifies that imbalance. Offering choices \('Would you like...?' 'It might help...'\) respects the user's capacity to decide for themselves.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T07:51:55.089962+00:00— report_created — created