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

Annotations already declare non-read-only, idempotent=true and destructive=false, but the description adds genuinely non-derivable behavior: completed screens are retained and still billed after a stop, and cancellation is asynchronous, requiring get_job polling until status is 'cancelled'. That billing/partial-work consequence is the kind of context annotations cannot express. Minor gap: no statement of what happens if the job already finished or was never started.

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?

Three short sentences, front-loaded with the action and its constraint, then the identifier convention, then the follow-up. The parenthetical '(agentJobId works too)' is slightly redundant with the schema's alias documentation, a small waste but not enough to disrupt scanning.

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 only two fully described optional parameters and no output schema, this is close to complete: the agent knows what the tool does, the identifier to pass, the billing consequence, and how to confirm cancellation. Absent is only edge-case handling for already-finished or unknown jobIds.

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 100% and both properties are documented, so the schema carries the parameter burden. The description restates that jobId is the agentJobId returned at start and that agentJobId is an accepted alias, which duplicates the schema text rather than adding new semantics. Baseline 3 is appropriate.

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?

The description names a specific verb and resource ('Stop an agent run') and scopes it to the three run kinds (design, React edit, video), with the state constraint 'that is still going'. It stops short of explicitly contrasting with the sibling 'delete' tool, though 'Finished screens stay and are billed' implicitly signals that this is a halt, not a cleanup.

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?

Use condition is clear: only for runs still in progress. It also hands the agent the correct follow-up ('Poll get_job until it says cancelled'), which is the key next-step guidance. It does not state what to do when the job is already finished, nor name an alternative tool such as delete for removing completed jobs.

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