Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare the safety profile (non-read-only, idempotent, non-destructive, open-world), but the description adds substantial behavior the annotations cannot: generation parameters are not inherited and must be re-passed, inputs expire after 2 hours, retries of retries are refused, the call spends workspace balance, and the call returns a run_id immediately while work continues in the background. This is exactly the kind of operational context that prevents misuse.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with purpose and scope before the operational caveats, and nearly every sentence carries distinct, non-redundant information. It is dense and long, but the length is justified by the number of distinct failure modes (expiry, double retry, billing, confirmation, async polling) it must cover.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With no output schema, the description still tells the agent what comes back (a run_id) and how to follow up (get_run with wait_seconds up to 50). For a 7-parameter, asynchronous, billed, confirmation-gated mutation, all the decision-relevant context an agent needs is present.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is only 57%, and the description compensates directly: it explains that temperature, reasoning_effort, web_search, and force_refresh are not remembered so the original values must be supplied, defines request_id as a per-run UUID whose reuse returns the same run without a second charge, and defines confirm_token as only valid after a needs_confirmation result. That is meaningful semantics well beyond the bare schema types.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource with precise scope: 'Run again only the rows of a finished run that failed, with the same prompt and model.' That scope immediately separates it from siblings like run_ai, run_flow, and get_run, which start new work or observe existing work rather than re-running failed rows.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives explicit when-to-use conditions (after a finished run with failures, within 2 hours while inputs are retained), explicit refusals (an already-retried run is refused; pass the newest retry's run_id), and named alternatives ('call estimate_cost first', 'call get_run with wait_seconds'). The needs_confirmation branch spells out the exact retry sequence with confirm_token.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.