Why can't an agent treat an approved status as permission to proceed?

**Why can't an agent treat APPROVED as permission to proceed?**

Because in many businesses the status field only records the decision, not whether its terms have been met. A record can read approved while three conditions remain open, and the systems that run later steps often check the status alone.

**Is this a rare edge case?**

No. Approvals with terms attached are routine wherever a decision-maker wants to commit early but control risk: lending, underwriting, procurement, patient transfers, supplier onboarding. The exact conditions vary, the structure does not.

**What does an agent have to do differently?**

- read the conditions, not just the status - track which are open, satisfied, waived or referred - tell apart evidence that arrives from evidence that passes - hold downstream actions until the gate opens - escalate when a condition cannot be met as written

**Why does this need generated data?**

Testing these behaviours needs many cases with known correct outcomes, including ones where the safe action is to wait. Generated records let a buyer construct those cases, with ground truth attached, without handling live customer files. The output is generated, so check how closely its structure and policies match your own process, and check for any unintended resemblance to real material before use.

**Does it replace testing on your own data?**

No. It gives you a controlled way to exercise gating behaviour early and repeatedly. Final validation in your own environment still matters.