Can generated environments include records that agents cannot retrieve?

**Can an environment include missing evidence?**

Yes. A release can represent evidence expected by the process but unavailable to the acting role. The important design choice is the reason for that absence and whether the agent can change it.

**Which distinctions are useful?**

A record can exist in an archive, exist behind an access restriction, have been removed under a retention rule, or never have been created. These conditions should lead to different responses. Requesting access may help with a restricted record but cannot recover an approval that was never recorded.

**How can the environment expose the difference?**

A search result might return an archive pointer, while an access request can return a permission decision. A retention register can explain removal. A process owner's response can confirm that an expected record was not created. The agent discovers the cause through permitted interactions rather than receiving a complete diagnosis at the start.

**What can a buyer request?**

Ask for an availability model, recovery actions and terminal outcomes for each absence type. Generated task labels can identify whether further action was possible. This lets builders test persistence without rewarding endless searches when the remaining gap requires an explicit exception.