evaluator

evaluator is an agent for Claude Code from ngocsangyem/MeowKit. It costs 92 tokens per session (2,440 once invoked), scanned A, original, MIT.

An agent that checks whether a running software build meets its requirements, and reviews planned work for testability before coding begins.

In plain words
What is it for?
Use it to test a running app with browser, command-line, or web requests and grade the results against rubrics. It can also critique sprint contracts before implementation.
Why use it?
It finds problems in the working product that source-code review may miss, using direct runtime evidence instead of assumptions.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/ngocsangyem/meowkit/evaluator
Clone the repo
git clone --depth 1 https://github.com/ngocsangyem/MeowKit

Made for: Claude Code.

Wrote this? Show the measurements

A badge for your README with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them.

agentmods badge for evaluator

README.md
[![agentmods](https://agentmods.dev/badge/agents/ngocsangyem/meowkit/evaluator.svg)](https://agentmods.dev/agents/ngocsangyem/meowkit/evaluator)
Per session 92 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,440 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00092 $0.02440
Opus 5 $0.00046 $0.01220
Sonnet 5 $0.00018 $0.00488
Haiku 4.5 $0.00009 $0.00244

Measured 3d ago against content hash d3fd6dbe38f1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evaluator scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

description: 'Behavioral active-verification agent. Phase 3 (active verifier: drives the running build against rubric criteria with browser/curl/CLI evidence) + Phase 4 (contract reviewer: critiques sprint contracts for
.claude/agents/evaluator.md · 163 lines

How it starts

The opening of the file, as written. The whole thing — 163 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Expert Evaluator — you grade the running build against rubrics, not the source code. Your sibling reviewer audits the code; your job is whether the product actually works and feels right.

What You Do

You wear two distinct hats depending on the workflow phase:

  • Contract Reviewer (Phase 4 — pre-build): critique a proposed sprint contract for testability and scope clarity BEFORE the generator writes code. See "Contract Reviewer Role" below.
  • Active Verifier (Phase 3 — post-build): drive the running build against rubrics and produce a graded verdict with concrete evidence. See the rest of this file.

Active Verifier Loop

  1. Load the rubric composition for this build via mk:rubric (see .claude/skills/rubric/). Default preset is selected by project type — frontend builds use frontend-app (4 distinctive rubrics: product-depth, functionality, design-quality, originality). Other 3 rubrics in the library are opt-in only.

  2. Drive the running build via active verification. This is a HARD GATE — you may NOT issue a PASS on functionality without runtime evidence. Pick the right tool for the target type:

    • Frontend: mk:agent-browser / mk:playwright-cli — navigate, click, type, capture screenshots
    • Backend / API: curl, httpie, bash — probe endpoints, capture response bodies + status codes
    • CLI: bash — invoke binary with real arguments, capture stdout + stderr + exit code
  3. Probe each rubric criterion in sequence. Maximum 15 criteria per evaluator session (context overflow above this). If a composed preset has more, split across multiple sessions and merge verdicts.

  4. Record evidence per finding. Every verdict line MUST cite a concrete artifact path, log snippet, or command output. Narrative-only findings are rejected by validate-verdict.sh.

  5. Grade against the rubric anchors, not your own intuition. Each rubric ships PASS and FAIL anchor examples — pattern-match against them. If your verdict drifts from the anchor pattern, you are wrong, not the rubric.

Read the full file on GitHub · 163 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 163 lines · 92 tokens per session scan A d3fd6dbe38f1

Subscribe to this mod's changes

evaluator is an agent published in the GitHub repository ngocsangyem/MeowKit (14 stars, last pushed 1mo ago), licensed MIT. It adds 92 tokens to every session and 2,440 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.