How does facility booking map onto environment state and tool actions?

A booking environment usually exposes a small set of read and write tools over linked tables.

**Read tools**

- `get_approval(approval_id)` returns amount, term, margin, repayment frequency and conditions. - `get_security(collateral_id)` returns the interest type and status. - `list_conditions(approval_id)` returns each condition with a status and evidence reference.

**Write tools**

- `create_facility(approval_id, terms)` creates a draft facility. - `link_security(facility_id, collateral_id)` attaches the security record. - `enable_drawdown(facility_id)` fails unless preconditions hold. - `generate_schedule(facility_id)` produces instalment rows. - `raise_exception(facility_id, reason_code, owner_role)` returns the file to a named role.

**State tables**

`approvals`, `conditions`, `facilities`, `security_interests`, `schedules`, `ledger_entries` and `exceptions`, joined by `approval_id`, `facility_id` and `collateral_id`.

**Reset**

Each episode starts from a seeded snapshot, so the same task id always produces the same starting records. Mid-episode events, such as an approval amendment, are scheduled by step number and replayed identically on reset.

The exact tool names and schemas are agreed during scoping, so they can match the harness you already run.