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

Annotations already declare readOnlyHint, openWorldHint, and idempotentHint, and there is no contradiction. The description adds substantial context beyond them: 'Cache TTL 12h' (staleness bound), 'Source: hand-curated from the upstream board at tbench.ai' (provenance), the per-view output shapes, and the non-obvious interpretation caveat that the same model can score in the 70s on 2.1 but the 20s on 4.0. This tells the agent how to interpret the data, not just that a read is safe.

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?

Dense but every sentence earns its place: function, row shape, interpretation guidance, comparability warning, per-view routing, source/TTL, and when-to-use. The most decision-critical facts (what it returns, the NOT-comparable caveat) are front-loaded, and the view routing is compressed into one efficient series. No filler or repetition.

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 one-optional-param read-only tool with annotations covering safety, the description is nearly complete: it explains row composition, view variants, freshness, provenance, and use cases. The gap is that no output schema exists and the description gives the conceptual row shape but not exact JSON field names or pagination behavior, so an agent invoking raw view would still have some discovery to do on response structure. Minor, but enough to hold it below a 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 coverage is 100%, but the schema only says 'Output shape; default raw' with enum values. The description adds real semantics for every enum value: summary = 'current-board ranking plus biggest harness gaps', gaps = 'full per-model harness deltas', combined = 'current board normalized to its top score', raw = 'full benchmark/result graph'. This goes well beyond the baseline 3 expected for full schema coverage.

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 first sentence states a specific verb and resource: 'Returns a snapshot of the public Terminal-Bench leaderboards, revisions 4.0 and 2.1.' It further clarifies the row structure (harness+model+95% CI), which distinguishes it from all 26 sibling tools — none of which handle leaderboards or benchmarks. The name 'tf_harnesses' alone is ambiguous, but the description fully disambiguates it.

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?

Ends with an explicit use case: 'Use when the agent needs to recommend a harness/model combo or explain why two agents using the same model perform differently.' It also embeds a critical exclusion — 'The two revisions are NOT comparable' — and routes the agent among the four view values for different needs. No competing sibling exists, so no alternative tool needs naming; the internal view routing fully covers selection.

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