04
Action space
Tools
Each turn the policy emits exactly one tool call. Malformed calls count as invalid queries.
| tool | signature | returns | guardrail |
|---|---|---|---|
| list_tables | () | Table names | — |
| describe_table | (table: str) | Columns and types | Unknown table → error |
| run_query | (sql: str) | Up to 200 rows | Read-only, 5 s timeout |
| submit_answer | (rows: list[list]) | Ends episode | Graded once |