improve-judge

improve-judge is a skill for Claude Code from experientiallabs/experiential. It costs 31 tokens per session (829 once invoked), scanned A, original, Apache-2.0.

A controlled process for improving a software judge that scores or classifies results, using reviewed disagreements and stored evaluation records.

In plain words
What is it for?
It is for reviewing scores, recording human corrections, recalibrating the judge, and approving changes in a router-based evaluation workflow.
Why use it?
It prevents score changes based only on intuition or unverified input and keeps new changes tied to a regression test, which checks that the same mistake does not return.

Skill for Claude Code

Written for Claude Code: installed under .claude/.

Good fit It is for reviewing scores, recording human corrections, recalibrating the judge, and approving changes in a router-based evaluation workflow.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/experientiallabs/experiential/improve-judge
About the project

Experiential is an open-source model gateway and router, meaning a service that gives agents one API for hosted, user-provided, local, and custom language models. It is for teams that need to choose models, control access and spending, and route production requests according to quality, speed, or cost. The catalogue entries provide agent skills and instructions for operating or configuring the gateway.

experientiallabs/experiential · 2,836 stars · on GitHub · experientiallabs.ai

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.

Any agent
npx skills add experientiallabs/experiential --skill improve-judge
Clone the repo
git clone --depth 1 https://github.com/experientiallabs/experiential

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 improve-judge

README.md
[![agentmods](https://agentmods.dev/badge/skills/experientiallabs/experiential/improve-judge/github.svg)](https://agentmods.dev/skills/experientiallabs/experiential/improve-judge)
Your own site
<a href="https://agentmods.dev/skills/experientiallabs/experiential/improve-judge"><img src="https://agentmods.dev/badge/skills/experientiallabs/experiential/improve-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 improve-judge

Your own site · 80×15
<a href="https://agentmods.dev/skills/experientiallabs/experiential/improve-judge"><img src="https://agentmods.dev/badge/skills/experientiallabs/experiential/improve-judge.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 31 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 829 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
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 $0.00031 $0.00829
Opus 5 $0.00015 $0.00415
Sonnet 5 $0.00006 $0.00166
Haiku 4.5 $0.00003 $0.00083

Measured 9d ago against content hash 449eb0001dd9, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

improve-judge 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 9d 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.

.claude/skills/improve-judge/SKILL.md · 81 lines

How it starts

The opening of the file, as written. The whole thing — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Improve the Judge

Use the common-owned judging contracts and persisted local review state. Every change starts from a reviewed disagreement in persisted rollout evidence and ends with a regression that would catch the same failure. Never tune a judge from intuition or from unverified caller data.

1. Resolve the canonical evidence

  • Start from one local project and its completed rollout, judgment, rubric, label, lineage, and calibration artifacts.
  • Use exp.common.judging.Judge as the scoring boundary. A judge receives recursively verified artifact IDs through judge_persisted and returns a structured Judgment.
  • Use RubricReview, HumanScoreReview, and JudgeCalibrationService to inspect judgments, record human score corrections, refresh calibration reports, and explicitly approve calibration. CLI and Platform workflows must call these same services rather than create a second artifact path.
  • Keep raw review and run output under the local project root or /tmp. Do not commit customer evidence or operator-local outputs.

For a composed router workflow, read the judgments persisted by exp.compose_router. The workflow injects its approved review supplier, setup supplier, simulator factory, judge, model catalog, and finite budgets. Do not create a second scoring or evaluation path around that composition seam.

2. Classify disagreements

Sample about 20 reviewed cells across the score range and compare every dimension judgment with the active human score. Classify each actionable miss:

  • A false positive scores materially above the human label.
  • A false negative scores materially below the human label.
  • A protocol failure lacks a valid structured Judgment and is infrastructure evidence, not a low score.
  • A lineage or provenance mismatch is an invalid input and must fail before calibration.

Recheck controls after every change. A new miss on an established control is a regression even if the target disagreement improves.

3. Locate the owning layer

Read the full file on GitHub · 81 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. 9d ago First seen · 81 lines · 31 tokens per session scan A 449eb0001dd9

Subscribe to this mod's changes

improve-judge is a skill published in the GitHub repository experientiallabs/experiential (2,836 stars, last pushed yesterday), licensed Apache-2.0. It adds 31 tokens to every session and 829 once invoked, about $0.0002 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 2026-08-30.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

local-ai-agents

Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…

microsoft/ai-agents-for-beginners · 200 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

insight-error-page

Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…

vercel/next.js · 83 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens

next-partial-prefetching-adoption

Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…

vercel/next.js · 103 tokens