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

Even with strong annotations (readOnlyHint, idempotentHint, non-destructiveHint), the description meaningfully adds depth: the disclosure that results are 'computed live', 'never asserted', and that an empty history returns null percentiles + sample_count 0 ('an honest empty') are material behaviors well beyond the structured metadata. The 'Verify, don't trust' line frames the appropriate mental model.

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 description is front-loaded and well-organized, but has clear redundancy and a weak closer ('Example: call federation_latency with arguments {}' adds no value beyond what the documented param pattern already conveys). The empty-history caveat ('an honest empty, not a fabricated number' / 're-call after traffic') appears nearly verbatim twice, which wastes words in what could be a tightly structured description.

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 properly documents the response shape '{ p50, p95, p99, sample_count }' and edge-case behavior (null percentiles for empty history). It also covers freshness/computation source and operational advice. Combined with strong annotations and a small parameter surface, the guidance is complete and digestible for the agent.

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?

The input schema already has 100% coverage of the two parameters, so the schema is doing the heavy lifting. The description does add some framing — 'scope to one tool' for `tool` and 'widen/narrow the sample window' for `limit` — which reinforns semantic meaning. However, it's largely reinforcing schema-provided context rather than unlocking semantic info unavailable elsewhere.

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 uses specific, precise language to define what's measured ('MEASURED p50/p95/p99 in ms of recent authenticated tool executions'), units, and the underlying data source (gateway's own audit records). It crisply differentiates from potential sibling observability tools by framing it as the 'verifiable answer to is it really sub-100ms?' — the grounding in 'real recorded call durations, never asserted' adds distinctive value.

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 description clearly conveys when to use it, including specific invocation patterns ('Omit args for the whole surface; tool= to scope; limit= to widen/narrow'), which provides implicit when-to-use guidance tied to use case. It stops short of a formal 5 because it never names alternative tools for when guidance would fall to another tool (e.g., 'use X instead for...'), though the rich usage examples largely compensate.

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