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

Annotations only declare readOnlyHint=false, destructiveHint=false and openWorldHint=true; the description adds substantial context beyond that: builds are queued asynchronously, each build of a `name` becomes the next version tagged latest (or `tags`), `start` controls sandbox runtime and readiness, a Dockerfile's CMD/HEALTHCHECK fills it in, and rebuilds reuse unchanged layers unless `cache` is false.

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?

The source-selection constraint and the async/versioning behavior are front-loaded, and every sentence carries information rather than filler. It is a dense block of many clauses in one paragraph, which slightly taxes parseability, but nothing is wasted.

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 16 parameters, nested objects, no output schema and only 75% schema coverage, the description supplies the missing lifecycle context: return value ('returns the image queued'), how to poll readiness, and how sandboxes consume the result by id or name:tag. A handful of auxiliary parameters remain unexplained, but the core workflow is complete.

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 75%, so the schema carries much of the load, but the description adds real meaning for the highest-risk parameters: the semantics of `recipe` fields, that COPY reads `files`, that `start` is what a sandbox runs at launch, and that `tags` defaults to latest. It does not document several declared params (`labels`, `region`, `dockerignore`, `buildArgs`, `build`), so it does not fully compensate for the coverage gap.

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 ('Build a sandbox image') and immediately enumerates the three mutually exclusive sources (`image`, `dockerfile`, `recipe`) with examples. It also names the related sibling (`runtime_images` action get) that reports readiness, so an agent can distinguish this build tool from the list/manage image tools without opening schemas.

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?

The 'exactly one of' constraint plus per-source descriptions make it clear which input to pick for a given situation, and it tells the agent where to check readiness afterward. It stops short of explicit when-not-to-use guidance (e.g., when a public image alone suffices and no build is needed), so it is clear context rather than full routing.

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