Video versus static advertising for a category, computed from how long real ads actually ran rather than from opinion. Returns a verdict, its confidence, and the ads behind it.
| Name | Required | Description | Default |
|---|---|---|---|
| category | Yes | The product category. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already provide readOnlyHint, idempotentHint, and destructiveHint, so safety is covered. The description adds meaningful behavioral context by stating the methodology (based on real ad run lengths, not opinion) and disclosing what is returned: a verdict, confidence, and supporting ads.
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?
Two sentences deliver the core purpose, the distinguishing methodology, and the output composition with no redundant wording. The most important information is front-loaded.
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 simple read-only, idempotent tool with one fully documented parameter, the description is complete. It explains what the tool computes, why it is reliable, and what the return will contain, which is especially helpful given there is no output schema.
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 the single 'category' parameter already described as 'The product category.' The description only repeats that the comparison is 'for a category' and does not add value such as allowed formats, examples, or edge cases. Baseline 3 is appropriate given 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 description clearly identifies the tool's focus: comparing video versus static advertising for a category and returning a verdict. It cites a specific resource (ad formats) and the underlying evidence, though it does not explicitly differentiate itself from sibling tools by name.
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 the tool is appropriate when an empirical, data-driven comparison is needed, since it is 'computed from how long real ads actually ran rather than from opinion.' However, it does not explicitly say when to use this tool over siblings like category_warmth or search_evidence, nor does it mention any conditions when it should not be used.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.
Each tool maps to a clearly distinct operation: checking category cache/warmth, comparing ad formats, resolving receipt IDs, and searching the evidence corpus. Even though search_evidence and get_receipt both return evidence, one is query-based and the other is a direct ID lookup, so there is no real ambiguity.
Three tools follow a clean verb_noun pattern: compare_formats, get_receipt, search_evidence. category_warmth is the one outlier as a noun phrase, but it is still descriptive and fits the domain's 'warmth' concept, so the inconsistency is minor.
Four tools is a compact but well-scoped set for an evidence-retrieval/corroboration server. Each tool earns its place: orient, search, compare, and verify. There is no redundant bulk or missing essential operation within the stated purpose.
The read-only evidence domain is covered end to end: category_warmth provides orientation, search_evidence returns corpus results with corroboration status, compare_formats delivers the comparison verdict, and get_receipt lets agents verify the underlying records. The workflow has no obvious dead ends.