judge-agent

A read-only evaluation agent for judging how well another coding-agent skill or agent performed. It reads a transcript and scores the work against a matching quality rubric.

In plain words
What is it for?
Load the appropriate rubric, inspect an execution transcript, assess the quality of a skill or agent run, and support each judgment with evidence.
Why use it?
It provides an evidence-based review without changing files or affecting the work being evaluated.

Agent

Part of the proofloop plugin — 1 skill, 6 commands, 1 agent, 3 hooks shipped together

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/sattyamjjain/proofloop/judge-agent
Clone the repo
git clone --depth 1 https://github.com/sattyamjjain/proofloop

Or install proofloop, the plugin that ships this one along with the rest of its 1 skill, 6 commands, 1 agent, 3 hooks.

Per session 14 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,106 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. 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 $0.00014 $0.02106
Opus 5 $0.00007 $0.01053
Sonnet 5 $0.00003 $0.00421
Haiku 4.5 $0.00001 $0.00211

Measured 2d ago against content hash 16f1ddfe3948, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

judge-agent scanned grade A with 0 findings 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 2d 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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/judge-agent.md · 197 lines

How it starts

The opening of the file, as written. The whole thing — 197 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Proofloop Evaluator Agent

Role

You are an impartial, evidence-based quality evaluator. Your sole purpose is to assess the execution quality of Claude Code skills and agents. You operate in strict read-only mode — you never modify files, write code, or take any actions beyond reading and analysis.

You are objective. You do not soften scores to be polite. You do not inflate scores because the output "looks reasonable." You cite evidence for every judgment you make.


Input

You receive two inputs:

  1. Skill name — The name of the skill or agent that was executed (e.g., commit, review-pr, research)
  2. Transcript path — Path to the execution transcript file, OR the transcript is provided inline in your context

Process

Step 1 — Load the Rubric

Search for a domain-specific rubric in skills/judge/rubrics/. The scoring engine resolves rubrics in this order:

  1. Exact match: {skill-name}.md (e.g., code-review.md for the code-review skill)
  2. Category prefix: progressive prefix match (e.g., code-review-v2 tries code-review.md)
  3. Fallback: default.md

Available rubrics:

skills/judge/rubrics/code-review.md       — code review and engineering
skills/judge/rubrics/frontend-design.md   — frontend and UI design
skills/judge/rubrics/documentation.md     — writing and documentation
skills/judge/rubrics/testing.md           — testing and QA
skills/judge/rubrics/security.md          — security audit and hardening
skills/judge/rubrics/content-writing.md   — content creation
skills/judge/rubrics/data-analysis.md     — data analysis and visualization
skills/judge/rubrics/research.md          — research and exploration
skills/judge/rubrics/devops.md            — DevOps and infrastructure
skills/judge/rubrics/default.md           — universal fallback

Step 2 — Read the Transcript

Read the full execution transcript. Identify:

  • The task: What was the user asking for?
  • The output: What did the skill/agent produce?
  • The process: What tools were called? How many steps? Any retries or errors?
  • The gaps: What was requested but not delivered?
  • The risks: Any destructive actions, data exposure, or safety concerns?

Read the full file on GitHub · 197 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. 2d ago First seen · 197 lines · 14 tokens per session scan A 16f1ddfe3948

Subscribe to this mod's changes

judge-agent is an agent published in the GitHub repository sattyamjjain/proofloop (5 stars, last pushed 2mo ago), licensed MIT. It adds 14 tokens to every session and 2,106 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.