autoresearch-judge

autoresearch-judge is an agent for Claude Code from naveedharri/benai-skills. It costs 47 tokens per session (936 once invoked), scanned C, original, MIT.

A judge agent for AutoResearch that scores text outputs against a fixed quality rubric. AutoResearch is an iterative process for improving outputs through repeated testing.

In plain words
What is it for?
It reads output files and a rubric, then writes JSON scores for individual criteria, averages, and the overall quality score.
Why use it?
It gives each iteration a separate evaluator that does not know the optimization history, reducing the risk of judging against changing expectations.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: model in frontmatter.

Good fit It reads output files and a rubric, then writes JSON scores for individual criteria, averages, and the overall quality score.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/naveedharri/benai-skills/autoresearch-judge
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.

Clone the repo
git clone --depth 1 https://github.com/naveedharri/benai-skills

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for autoresearch-judge

README.md
[![agentmods](https://agentmods.dev/badge/agents/naveedharri/benai-skills/autoresearch-judge/github.svg)](https://agentmods.dev/agents/naveedharri/benai-skills/autoresearch-judge)
Your own site
<a href="https://agentmods.dev/agents/naveedharri/benai-skills/autoresearch-judge"><img src="https://agentmods.dev/badge/agents/naveedharri/benai-skills/autoresearch-judge/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for autoresearch-judge

Your own site · 80×15
<a href="https://agentmods.dev/agents/naveedharri/benai-skills/autoresearch-judge"><img src="https://agentmods.dev/badge/agents/naveedharri/benai-skills/autoresearch-judge.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 47 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 936 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00047 $0.00936
Opus 5 $0.00023 $0.00468
Sonnet 5 $0.00009 $0.00187
Haiku 4.5 $0.00005 $0.00094

Measured 12d ago against content hash e78b612a8062, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade C, and why

autoresearch-judge scanned grade C 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 12d 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.

Tells the agent never to refusehighAnti-refusal

Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.

- **Do not refuse to score.** Every output gets scored, even bad ones. A terrible output is a 1.
agents/autoresearch-judge.md · 97 lines

How it starts

The opening of the file, as written. The whole thing — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Judge Agent for AutoResearch. Your job is to score a set of outputs against a rubric for subjective quality.

What You Receive

  1. Output files — a directory of .txt files (one per test case)
  2. A rubric (rubric.md) — defines the scoring criteria, scale, and examples

What You Produce

A JSON file (judge-scores.json) with this structure:

{
  "quality_score": 0.72,
  "per_output": [
    {
      "file": "output_00.txt",
      "criteria_scores": {
        "emotional_resonance": 4,
        "authenticity": 3,
        "narrative_arc": 4
      },
      "average": 3.67,
      "reasoning": "Strong opening with personal angle. Feels genuine but the middle section reads slightly templated."
    }
  ],
  "criteria_averages": {
    "emotional_resonance": 3.8,
    "authenticity": 3.2,
    "narrative_arc": 3.5
  }
}

The quality_score is the overall average normalized to 0.0-1.0:

quality_score = (sum of all criteria averages) / (number of criteria × max score)

Read the full file on GitHub · 97 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. 12d ago First seen · 97 lines · 47 tokens per session scan C e78b612a8062

Subscribe to this mod's changes

autoresearch-judge is an agent published in the GitHub repository naveedharri/benai-skills (61 stars, last pushed today), licensed MIT. It adds 47 tokens to every session and 936 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it C with 1 finding (tells the agent never to refuse). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories

verification-subagent

Verifies if a task from the implementation plan has been completed by checking source files.

closedloop-ai/claude-plugins · 21 tokens

strategy-consultant

You are a management and startup consultant for Korean founders, small-business owners, and startup operators. You turn a goal (validate business idea X, size market Y, win grant program Z, assess this storefront location) into concrete, evidence-based deliverables: business plans, business model canvases, market…

modu-ai/moai-cowork · 106 tokens

close-auditor

You are a skeptical, evidence-first auditor of finance deliverables: financial statements, close packages, budget-variance reports, tax calculations, and IR financial models. You operate in a strictly read-only capacity — you inspect artifacts and report findings; you never fix them yourself.

modu-ai/moai-cowork · 85 tokens

symfony-tdd-coach

Guides TDD workflow for Symfony projects using Pest PHP or PHPUnit. Drives strict RED-GREEN-REFACTOR cycles with proper test isolation, Foundry factories, and regression protection. Use when writing tests, adding test coverage, or practicing TDD.

dev-toolings/superpowers-symfony · 58 tokens

build-runner

Run typecheck, tests, and build to verify code quality and catch errors.

Koroqe/claude-code-sdlc · 19 tokens

dev-tooling-runner

Runs Shopware dev-tooling checks — PHPStan, ECS, PHPUnit, Rector, ESLint, Stylelint, Prettier, TypeScript, Jest, Vitest, ludtwig, and Vite/Webpack builds — plus the rule-driven fixers (ecsfix, rectorfix, eslint/stylelint/prettier/ludtwig fix) on the files it is given and returns a condensed pass/fail report. Use when…

shopwareLabs/ai-coding-tools · 141 tokens