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

Annotations only say readOnlyHint=false and destructiveHint=true; the description goes well beyond by disclosing that answers are persisted, that unreview retracts an answer, that only reviewed rows train, and that use_reviewed overwrites the model's training data. It also flags that confidence numbers are AMT's own and must not be recalculated, and warns about instructions embedded in row text/labels.

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 core read-and-record action is front-loaded, and the remaining sentences cover workflow consequences and two safety constraints. It is dense rather than padded, though the number-handling and injection warnings add length that a shorter phrasing could partly absorb.

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 an 8-parameter destructive mutation with no output schema, the description supplies the workflow context, mutation consequences, and trust boundaries an agent needs. It leaves the filter enum values and offset behavior to the schema and says nothing about the result payload, which keeps it short of 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?

With only 38% schema description coverage across 8 parameters, the description carries real weight: it explains that limit is capped at 100 with least-sure rows first, and defines the semantics of answers (confirm or change), accept_above (accept at or above a confidence), and unreview (take an answer back). filter modes and offset are still only inferable from the schema, so it does not fully close 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?

The description gives a precise verb+resource: it reads a cold start's rows together with AMT's suggested category and confidence, and records the user's answers. It also situates the tool in the workflow (only reviewed rows train; use_reviewed then train), which separates it from siblings like train and use_reviewed.

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?

It states the operating modes clearly (confirm/change rows via answers, bulk-accept via accept_above, undo via unreview) and explains the downstream condition for use_reviewed/train. It does not explicitly contrast with nearby siblings such as correct_labels or suggest_labels, so the routing guidance is strong but not exhaustive.

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