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

Adds substantial behavior beyond the annotations: nothing persists, failures come back as repair-grade problems rather than partial state, the call takes ~15s, the response exposes computer.sites and whether the Task permits each site, and there is a confirmation path. The 'nothing persists' claim is a conservative but not contradictory pairing with readOnlyHint=false, which is based on the compile being a non-trivial compute rather than a pure read.

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-loads the highest-value facts — what it does and that nothing persists — then the failure mode and latency. Dense but nearly every sentence earns its place; the presentation instruction ('which clauses run flows, which you will handle, who decides') is the one line that could be trimmed.

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 carries the burden of describing returns and does so at a usable level: a proposed plan, repair-grade problems, and the computer.sites field. The finer shape of the plan object is left unspecified, but an agent knows what comes back and what to do with it.

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%, so taskId, sentence, workspace and approvalId are already documented. The description only adds meaning around the sentence (it must be refined iteratively) and doesn't touch taskId or workspace at all, so it sits at the baseline for well-documented params.

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 — compile a sentence into a PROPOSED Task plan — and makes the non-persisting, proposal-only nature explicit. That distinguishes it from workbench_tasks_create, workbench_tasks_freeze and workbench_tasks_set_sites, all of which it names or implies as separate steps.

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 concrete usage rules: refine the sentence with the user rather than guessing when the compiler returns problems, use workbench_tasks_set_sites to resolve disallowed sites, then iterate or freeze. It stops short of stating when to reach for compile versus get/update on an existing task, so not a full when/when-not map.

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