Why would a model lab want material about facility setup?

**Why would a model lab want material about facility setup?** Because it is a long-horizon task with partial information. The agent reads an approval, an offer, identity and security records, then writes to several systems in a required order. Mistakes often show up later, which makes them hard to catch with short-horizon tests.

**Is this just form filling?** No. The data entry is the easy part. The hard parts are conditions precedent that are met in unusual ways, entities that appear under different names, and handovers where the next team assumes the previous one checked something.

**What can a generated release add that public text does not?** Linked state: an approval, an offer, a facility, a schedule, a security record and an audit trail that all agree or deliberately disagree. Public text rarely shows system state before and after each action.

**Does generated mean safe to use without checking?** No. The material is generated, not drawn from real records, but a buyer should still inspect how closely it follows real practice and whether any pattern could resemble an identifiable organisation. Ask for that inspection in a sample.

**Which roles appear?** Credit officer, settlement officer, compliance reviewer, servicing team lead and sometimes a relationship manager. Each holds different permissions, which makes escalation behaviour testable.