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

Annotations state readOnlyHint=false, openWorldHint=true, and destructiveHint=false. The description adds behavioral context: it notes that the wrapper may not reliably produce a structured reproduction bundle, and that output is a high-level note rather than a proof. This is useful transparency about reliability and output nature, beyond what annotations provide. However, it doesn't disclose potential side effects (though readOnlyHint=false suggests possible writes), and given the experimental nature, more disclosure could be expected. The description does not contradict annotations, so a 3 is appropriate.

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 concise, with three sentences packed with information: purpose, experimental caveat, output quality, and pricing. It is front-loaded with the core function. However, the last phrase 'a tool · audit' is cryptic and adds little value, and the pricing information might be better in annotations or separate metadata. Slight points deducted for that noise.

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?

Given the complexity (experimental, one nested object parameter, no output schema), the description clarifies expected delivery (counterexample or reason) and limitations (not machine-checkable), which is critical for setting agent expectations. It does not describe the exact input format beyond the schema, but the schema is self-explanatory with examples. It is complete enough for an agent to decide whether to use it, though a bit more detail on input could push it to 5.

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 description coverage is 100%, with a single 'input' object parameter described as a capability-specific payload with examples. The description mentions giving it 'one code claim plus relevant code' but does not detail the exact structure of the input object beyond the schema's generic example. Since the schema already covers the parameter, the baseline is 3, and the description adds a bit of context about what to pass, but not enough to raise it.

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 clearly states the tool's purpose: it takes a code claim and relevant code to produce a counterexample or reason the claim fails. It identifies the resource (code claim) and verb (refute). It is distinguished from siblings by being experimental and an 'analyst's pass,' not a machine-checkable proof, which sets it apart from more formal tools like code-review. However, it does not explicitly name which sibling to use instead, so it loses a point.

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 implies when to use: when you have a code claim to refute, and it hints at limitations (not a machine-checkable proof). It does not explicitly state when not to use it or alternative tools, but the context of 'experimental' and the caveat about the output quality give clear usage context. It is better than average but not fully explicit.

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