Stalls instead of acting

“The AI stalls instead of acting”

You authorized the work, and you got a plan, a recap, and a question that wasn’t one. These techniques remove the stall at its source — the slot where permission-seeking lives.

Which version of it is yours?

• Action-First When Clear, Ask When Unclear — it narrates the plan and waits for a nod, or guesses when it should ask
• Caution-as-Evasion Loop — three confirmations deep, every reply is ‘yes, proceed’, and the task still hasn’t moved