candidates

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

With zero annotations, the description carries the entire behavioral burden — and it does so well. It discloses that candidates are never suppressed based on constant size, nearby guards, or missing input flow; it states the cost model ('Costs one catalog bind on first call per graph (cached after)'); and it explains the coverage semantics of the 'frontiers' output (counts vs the census's full roster). It does not disclose pagination or what happens on empty results, but the core safety and cost behavior is transparently declared.

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 tight: a clear opening sentence, a sharp POINTER vs safety-check clarification, and then three compact operational facts (first constructor, cost/caching, frontier semantics). Every sentence earns its place with no filler. It reads like a well-edited paragraph, though grouping the three operational notes into a short list would make scanning even faster.

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 ranking tool with no output schema and 7 parameters, this is only partially complete: it defines behavior, cost, and the closest sibling, but it does not explain pagination via cursor/limit, the language/domain/detail value space, or what the output should look like for detail modes. The 'frontiers' note adds output hints, but given no output schema and the large sibling set with similar-sounding tools (scan, coverage_map, candidate_census), a bit more structural guidance on what a result page contains would round it out.

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 only 29%, and the description adds almost no parameter-level meaning — only a stray hint that 'the first constructor is memory.copy.' for capacity', which maps to constructor_id. limit, cursor, domain, and language are left undocumented in both the schema and description, and the very shallow phrase 'candidates are never suppressed...' implies the domain is a fixed graph already selected by a prior 'Atropos fact', so the agent cannot tell what one would pass for domain or language. The description does not compensate for the large schema coverage gap.

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 gives a specific verb and resource — 'Enumerate and rank every observable obligation site selected by an Atropos fact' — and immediately clarifies the tool's role as a POINTER, not a safety check, which separates it from analysis or validation tools. It does not fully spell out what an 'obligation site' or 'Atropos fact' physically means, so an agent must infer some domain vocabulary, but the core enumeration-and-rank purpose is unmistakable.

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 gives explicit routing to a sibling: 'call candidate_census for the full roster of catalog sinks that never bound', which is an explicit when-not-to-use signal, and the 'POINTER, not a safety check' line tells the agent it is inappropriate for validation workflows. However, it never distinguishes when to reach candidates versus the closer siblings like candidate_detail (drill-down) or coverage_map, so the alternative mapping is only partial.

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