Skip to main content
Glama

Check a rubric against human labels

check_against_humans

Run a rubric over human-judged gold sets to measure pass/fail and per-criterion agreement, flag low-confidence escalations, and preview each disagreement.

Instructions

Run a rubric over a gold set (subjects people already judged) and report how often Verdict reached the same verdict: overall pass/fail agreement, per-criterion agreement, how many subjects the judge handed to a human (low-confidence escalation), and every disagreement with a preview. rubricPath is a local YAML/JSON rubric file, re-read on every call; gold is a local JSON array or JSON-lines file of {id?, name, inputs, expected?: {criterion: label}, human?: {passed?, criteria?}, labelSource?, split?}; choice criteria are labelled with an option id, and unreadable labels are rejected before any judge runs. Local mode only.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "gold": "<gold>"
    },
    "name": "check_against_humans"
  }
}

Input Schema

NameRequiredDescriptionDefault
goldYesLocal gold-set file: JSON array or JSON lines
rubricNoStored rubric id or inline rubric, instead of rubricPath
rubricPathNoLocal rubric file (YAML or JSON)

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

A4.1/5.0
Behavior4/5

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

No annotations are supplied, so the description carries the full burden and does a good job: it discloses that the rubric file is re-read on every call, that unreadable categorical labels are rejected before any judge runs, that a human-escalation count is part of the report, and that the tool is local-mode only. It stops short of saying whether a judge run is costly/long-running or what happens on partial gold sets.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with the action and its outputs, then parameter detail, then the local-mode constraint. Dense but nearly every clause adds a fact; the nested-brace gold schema inline makes it heavier to read than ideal.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With no output schema and no annotations, the description still names the report's contents and the input file formats, which is what an agent needs. Remaining gaps are runtime expectations (cost of a judge pass, timeouts) and the undocumented `rubric` argument.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

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 extra by specifying the gold file's inner shape ({id?, name, inputs, expected?, human?, labelSource?, split?}) and the rule that choice criteria are labelled by option id — details the schema does not contain. The `rubric` parameter (stored id or inline rubric) is left undocumented in prose.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Names a specific verb and resource: run a rubric over a gold set of human-judged subjects and report agreement. It also enumerates the concrete outputs (overall pass/fail agreement, per-criterion agreement, escalation counts, disagreements with preview), so an agent knows exactly what this produces and how it differs from a generic evaluate or calibration_report.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The use case is implied clearly (measuring judge agreement with human labels before trusting a rubric), and 'Local mode only' plus the file-based inputs bound when it applies. However, it never states when to pick a sibling such as calibration_report or compare instead, nor any prerequisite ordering.

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