auto improve criterion

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

Annotations already declare readOnlyHint=false and idempotentHint=false, so the description's real contribution is cost and side-effect disclosure: 'SPENDS MONEY' and 'one metered rewriter call (billed under assist:iterate) plus the queued alignment run'. It also exposes partial-failure behavior (nothing created, only the rewriter call spent) and the deliberate single-round design, which no annotation conveys. Nothing contradicts the annotation hints.

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 definition is long but front-loaded: the first sentence carries purpose, mechanism, and when-to-use. The exhaustive return-field list earns its place only because no output schema exists, but the billing warning is stated twice ('SPENDS MONEY' and 'SPENDS THE WALLET'), which is redundant and could be tightened.

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?

For a money-spending, non-idempotent, queueing operation with one required parameter and no output schema, everything an agent needs is present: return shape (201 with full criterion object), error conditions (404/400 with their triggers), cost, auth scope (evals:write), post-call expectations (loop yourself; adoption stays human), and maxDuration. No material gap remains.

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 the single parameter already has a descriptive schema entry ('Criterion id (must belong to the key's workspace)'), so the baseline is 3. The description only restates the parameter's role via the endpoint path criteria/{id} and adds no new semantic meaning beyond what the schema provides.

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 opens with a specific verb and resource: 'Runs one auto-improvement round on a judge criterion' and decomposes it into three concrete steps (mines tune-half disagreements, rewrites the judge prompt,creates a successor DRAFT criterion with alignment queued). This clearly differentiates it from siblings like run_criterion_alignment and create_criterion by spelling out what auto-improve means as a distinct post-alignment operation.

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?

Explicit when-to-use guidance is given: 'use it when a calibrated judge still disagrees with your grades and you want a better candidate without hand-editing the prompt.' When-not conditions are enumerated as 400 triggers (alignment in progress, fewer than 80 judged rows, zero disagreements, unusable rewriter), and the 'Deliberately single-round: loop it yourself' note plus 'adoption stays a human act' tells the agent how to orchestrate multi-round use.

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