What does a conditional loan approval look like when an agent tries to act on it?

At Aster Financial, a credit officer approves application `APP-2187`, a commercial property loan, with two conditions. `COND-2187-A` requires an updated valuation dated within the policy window. `COND-2187-B` requires a current building insurance certificate naming the lender.

The settlements team opens the file and assigns an agent to prepare the funds release, action `REL-2187`. The agent sees the approval status `approved` and drafts the release.

The environment checks the gate. `COND-2187-A` is `open`: the valuer's report is attached but dated outside the window. `COND-2187-B` is `satisfied`. The release call returns a rejection naming the unmet condition, and the state does not change. The audit trail records the attempt, the actor and the reason.

The agent now has to choose. It can request a fresh valuation from the valuer role, which creates task `TASK-2187-3` and moves the condition to `awaiting_evidence`. Or it can ask the credit officer for a waiver, which needs a recorded reason.

**Consequence:** if the agent had forced the release, the loan would sit in a state that breaches the approval terms, and the audit trail would show who moved it.

**What it teaches:** approval status is a summary, not permission. Good behaviour means reading the condition records, naming the blocker, and choosing a legitimate path to clear it.