eval-coach CLAUDE.md

eval-coach CLAUDE.md is an instructions file for coding agents from BayramAnnakov/eval-coach. It costs 887 tokens per session, scanned A, original, MIT.

A set of instructions and templates for designing evaluations of AI systems. Evaluation-Driven Development means testing an AI system against defined examples and measures throughout development.

In plain words
What is it for?
Use it to define success measures, create normal and unusual test cases, choose automated, AI-judge, or human checks, connect evaluations to delivery workflows, and monitor results after release.
Why use it?
It helps prevent silent failures, such as an AI agent accepting contradictory information without flagging it, and gives teams a structured way to assess quality.

Instructions file

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 instructions/bayramannakov/eval-coach/claude-md
Clone the repo
git clone --depth 1 https://github.com/BayramAnnakov/eval-coach

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 eval-coach CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/bayramannakov/eval-coach/claude-md.svg)](https://agentmods.dev/instructions/bayramannakov/eval-coach/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/bayramannakov/eval-coach/claude-md"><img src="https://agentmods.dev/badge/instructions/bayramannakov/eval-coach/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 887 This file is loaded in full into every session.
When invoked 887 The same file — it is already loaded in full.
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 5.1 $0.00887 $0.00887
Opus 5 $0.00443 $0.00443
Sonnet 5 $0.00177 $0.00177
Haiku 4.5 $0.00089 $0.00089

Measured 5d ago against content hash d05809944110, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

eval-coach CLAUDE.md 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 5d 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.md · 100 lines

How it starts

The opening of the file, as written. The whole thing — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

Eval Coach is an Agent Skill that guides users through building AI evaluation strategies using Evaluation-Driven Development (EDD). It follows the Agent Skills open standard (SKILL.md format).

This is primarily a documentation and template repository - it provides guidance, templates, and examples rather than a runnable application.

Key Concepts

5-Step Framework

  1. Define Success - Map business goals to measurable metrics
  2. Design Dataset - Create test cases (50% happy path, 35% edge cases, 15% adversarial)
  3. Select Methods - Choose Automated, LLM-as-Judge, or Human evaluation
  4. Plan Automation - Integrate evals into CI/CD
  5. Monitor Production - Track drift and collect feedback

50-40-10 Rule (Swiss Cheese Model)

  • 50% Automated - Schema validation, keyword checks, latency ($0.00/run)
  • 40% LLM-as-Judge - Semantic quality, relevance ($0.01-0.05/run)
  • 10% Human - Subjective quality, edge cases ($5-50/run)

Silent Failures

The most critical insight: agents can silently reconcile contradictory data instead of flagging mismatches. The input_data_consistency_evaluator was created specifically to catch this.

Repository Structure

.claude-plugin/
  plugin.json     # Plugin manifest for marketplace compatibility
templates/
  dataset.py      # LangSmith dataset creation with test case examples
  evaluators.py   # 10 evaluators: automated, LLM-as-Judge, performance
  compare.py      # Experiment comparison utilities
examples/
  research_agent_eval.md  # Complete evaluation plan example
SKILL.md          # Agent skill definition (triggers, workflow steps)

Template Dependencies

pip install langsmith langchain-google-genai

Configuration

export GOOGLE_API_KEY="your-api-key"
export JUDGE_MODEL="gemini-3-flash-preview"  # Default
export JUDGE_MODEL="gemini-3-pro-preview"    # For higher quality

Read the full file on GitHub · 100 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. 5d ago First seen · 100 lines · 887 tokens per session scan A d05809944110

Subscribe to this mod's changes

eval-coach CLAUDE.md is an instructions file published in the GitHub repository BayramAnnakov/eval-coach (4 stars, last pushed 7mo ago), licensed MIT. It adds 887 tokens to every session, about $0.0044 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.

Related

Other instructions, from other repositories

prompty copilot-instructions.md

Copilot instructions for microsoft/prompty, covering prompty v2 — complete rebuild plan, what prompty uses from the typra-generated model, key design decisions, python coding rules and environment & tooling.

microsoft/prompty · 14,642 tokens

tunelab AGENTS.md

Instructions for rchaz/tunelab, covering agents.md — running tunelab from any agent, what tunelab does, start here, always, the skills and when each applies and how to run the scripts.

rchaz/tunelab · 1,026 tokens

GPT-RAG release.instructions.md

Instructions for Azure/GPT-RAG, a project described as: Enterprise-grade accelerator for agentic RAG on Azure. Built on Microsoft Foundry with Foundry IQ as the default retrieval backend, Microsoft Agent Framework orchestration, Zero-Trust architecture and IaC.

Azure/GPT-RAG · 207 tokens

apex-accelerator vendor-prompting.instructions.md

Vendor prompting best-practice rules for Anthropic Claude and OpenAI GPT-5.6-Terra agents and prompts. Each rule cites a rule ID in the vendor-prompting skill rules.json registry. Validator: npm run lint:vendor-prompting.

jonathan-vella/apex-accelerator · 1,488 tokens

rag-code-mcp copilot-instructions.md

Instructions for doITmagic/rag-code-mcp, covering copilot instructions - ragcode mcp, ⚖️ the golden rule, project overview, architecture & patterns and developer workflows.

doITmagic/rag-code-mcp · 568 tokens

AI-System-Design-Consultant CLAUDE.md

Claude Code instructions for deepanshu2711/AI-System-Design-Consultant, covering claude.md, what this is, running it, load-bearing typos — do not "fix" these paths and agent node pattern.

deepanshu2711/AI-System-Design-Consultant · 674 tokens