Cognates in parallel

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

Annotations declare readOnlyHint=true and the description is consistent with a read-only surface, even warning about not leaking restricted material. It adds real behavioral context beyond the annotations: license fields must be preserved when quoting, the `borrowed` flag can be null and needs shelf-language judgment, and recall is bounded by Wiktionary and gold-lemmatization coverage. However, the description is aimed mostly at result interpretation rather than operational behavior (latency, caching, or errors beyond the lect-module note), so it sits at a solid 3 rather than higher.

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?

It is front-loaded with the core definition and example before the mechanics, and every clause is doing work — licenses, borrowed flags, suppression, recall bounds. It is dense and jargon-heavy ('shelf-visited multi-hop walk', 'SHELF language'), which costs a point, but there is little pure filler.

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?

There is no output schema, so the description carries the return-shape burden and does: each group carries the verse ref, the root (headword, shelf language, gloss, license) and per-language witness words (lemma, surface forms, attesting documents with licenses, borrowed flag). Combined with the recall caveats and the bounded default/max, an agent has enough to call this and interpret the results correctly.

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%, so the baseline is 3, but the description earns an extra point by explaining intent the schema does not: `target` as a work id batches the whole work, `langs` must name at least two languages and defaults to every gold language, and `all` exists specifically to defeat the frequency-proxy suppression. It does not add much on `lect` or `include_restricted` beyond what the schema already says.

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 names a specific operation over a specific resource ('cognates-in-parallel over the local nabu corpus: verses ... where witnesses in TWO OR MORE languages use reflexes of the SAME reconstruction root') and gives a concrete example (Gothic salt ~ OCS соль). It also distinguishes itself from siblings by naming nabu_align as the verse-column supplier and nabu_etym as the root-closure supplier, so an agent knows this is the join tool, not either component.

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?

Usage is clear: it compares two-or-more languages, `langs` narrows to named languages, `target` is a work or a ref, and `all` lifts common-word suppression. The description also states a genuine limitation ('no hit is absence of evidence, not evidence of unrelatedness') that governs how to interpret results. It does not explicitly contrast with nabu_parallels or say when to prefer nabu_etym alone, which keeps it below a 5.

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