Verify Korean statute and case citations in legal text by checking if each cited law, article, and precedent exists as of a date and whether titles and claims match.
글 속 법령 조문·판례 인용을 뽑아 실제로 존재하는지(기준일 판), 제목·내용이 맞는지 판정한다. 법령명은 사전 일치로만 잡고, 업스트림 장애는 UNCHECKED(환각 판정 아님).
| Name | Required | Description | Default |
|---|---|---|---|
| asOf | No | 기준일(YYYY-MM-DD, 기본 오늘) | |
| text | No | ||
| citations | No | ||
| threshold | No | ||
| checkContent | No |
Changes observed during successful MCP inspections.
v0.1.0Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
readOnlyHint=true already tells the agent this is a safe read, and the description adds genuinely non-obvious behavior: law names are matched only via dictionary (so a name miss is not proof of hallucination) and upstream failures yield UNCHECKED rather than a hallucination verdict. That distinction materially changes how an agent interprets results. It still omits rate limits and whether asOf affects cached data.
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 dense sentences with no filler, and the core capability leads before the caveats. The parenthetical clarifications (UNCHECKED, dictionary match) are compact and earn their place, though the phrasing is compressed enough that the two model modes (text vs citations) never get separated.
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?
With no output schema, the description does outline the verdict space (exists / title-content mismatch / UNCHECKED), which is the most important return semantics. However, with five parameters at 20% schema coverage and no output schema, the absence of any guidance on text-vs-citations inputs, threshold, and checkContent leaves real gaps for correct invocation.
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 20% – only asOf is documented – so the description carries the burden and largely drops it. It hints at the law-name matching rule relevant to citations.law and repeats the base-date concept already in the schema, but says nothing about text vs citations, what threshold (0.5–1.0) controls, or what checkContent toggles.
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?
States a concrete verb+resource pair: extract statutory/case citations from text and judge whether they actually exist and whether title/content match. That is far more specific than the bare name citation_audit, though it never names a sibling to contrast with (statute_find, precedent_trace) so the agent must infer the boundary itself.
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 implied rather than stated: 'audit' plus the existence/match judging makes the verify-my-citations case clear, but there is no explicit when-to-use or when-not, and no routing to the sibling lookup tools (statute_find, article_get) that an agent would otherwise consider for similar needs.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.