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

Annotations already declare this as a safe, read-only, idempotent operation, and the description goes well beyond that with non-obvious traits: shared-VM measurement variance (the 12.5s / 4.7s / 12.7s LCP example), the absence of field data for low-traffic pages, the ~90-second cost, and the fact that a cached PageSpeed answer refunds the extra call and is flagged in the reply.

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?

Purpose and outputs are front-loaded in the first sentence, and the runs paragraph is placed after them. The anecdotal LCP figures are slightly verbose but they are the evidence that justifies the runs parameter, so they earn their place; the block is dense rather than bloated.

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 pay-per-call, slow, high-variance external audit the description is nearly complete, covering cost, timing, variance, and refund behavior. It does spend several lines restating return fields that the output schema already defines, and it omits any hint about how to choose among the other audit/technical-audit siblings.

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 0%, so the description must carry the load, and it does: runs=2 is defined precisely (a second measurement ~1 minute later, reported alongside median and spread under 'stability') and strategy's allowed values mobile|desktop are given even though the schema has no enum. url itself is left self-evident.

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 ('AUDIT of a web page') and enumerates the exact artifacts produced: Lighthouse scores, LCP/CLS/INP, failed audits with failing elements, render-blocking CSS, and loaded measurement tags. It does not, however, differentiate itself from closely related siblings such as core_web_vitals, seo_technical_audit, or instant_audit, which an agent could easily confuse it with.

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 clear and explicit operational guidance: use runs=2 before and after a speed change to separate signal from noise, and explains that lab numbers on the shared VM move while field data does not. It never states when to prefer this tool over an alternative sibling, so the routing guidance is one-sided.

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