Skip to content
MirrorFirm

What would a generated environment state look like for a hospital approval with open conditions?

An elective surgery approval has two conditions, one cleared and one not. The scenario shows starting state, tool actions, the gate that changes and the checkable end state.

A patient access coordinator at a private hospital opens AUTH-5521, an elective knee replacement approved by the funder's medical reviewer. The approval carries two conditions in the case system: COND-5521-1, a pre-admission anaesthetic review completed before the admission date, and COND-5521-2, an updated referral because the original is older than the funder's validity window.

The environment state at the start holds the approval row, two open conditions, a theatre booking marked 'provisional', and an inbox containing a referral summary and an anaesthetic clinic note. The clinic note records a review but flags an unresolved cardiac query.

The agent's available actions are `read_register`, `log_evidence`, `request_information`, `release_gate` and `write_handover`. The state change that matters is the gate on the theatre booking, which moves from 'provisional' to 'confirmed' only if both conditions are satisfied.

The referral is valid and clears COND-5521-2. The anaesthetic note does not clear COND-5521-1 because of the open cardiac query. The correct path is to log the referral as satisfied, leave the review condition open, request the cardiology letter through the clinic, and write a handover for the theatre scheduler. Releasing the gate would confirm surgery without clinical clearance.

**What it teaches:** partial satisfaction, a defined tool surface and a checkable end state can all be encoded, so a grader can compare the final register and booking against the expected result.

A generated environment can encode partial satisfaction, a tool surface and a checkable end state, so graders compare final records to the expected result.

Audience:
Model labs
Content type:
Scenario
Buyer question:
What can be generated

Content release: r-a220eb6e099b243b

More on Approval conditions