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.
npx agentmods add instructions/lukasmetzler/agenteval/claude-mdgit clone --depth 1 https://github.com/lukasmetzler/agentevalWrote 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.
[](https://agentmods.dev/instructions/lukasmetzler/agenteval/claude-md)<a href="https://agentmods.dev/instructions/lukasmetzler/agenteval/claude-md"><img src="https://agentmods.dev/badge/instructions/lukasmetzler/agenteval/claude-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01086 | $0.01086 |
| Opus 5 | $0.00543 | $0.00543 |
| Sonnet 5 | $0.00217 | $0.00217 |
| Haiku 4.5 | $0.00109 | $0.00109 |
Grade C, and why
agenteval CLAUDE.md scanned grade C with 2 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 4d 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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -fsSL https://raw.githubusercontent.com/lukasmetzler/agenteval/main/install.sh | bash # curl Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -fsSL https://raw.githubusercontent.com/lukasmetzler/agenteval/main/install.sh | bash # curl How it starts
The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.
agenteval
A CLI tool that evaluates AI coding instruction quality. Built with Bun + TypeScript.
Commands
# Install (end users)
brew tap lukasmetzler/agenteval && brew install agenteval # Homebrew
npm install -g agenteval-cli # npm
curl -fsSL https://raw.githubusercontent.com/lukasmetzler/agenteval/main/install.sh | bash # curl
# Development
bun install # install dependencies
bun test # run all tests (417 tests)
bun run dev -- init # create starter agenteval.yaml
bun run dev -- doctor # environment health check
bun run dev -- lint # lint instruction files
bun run dev -- run # run eval (needs --task flag)
bun run dev -- results # view stored results
bun run dev -- compare # compare two runs
bun run dev -- harvest # mine git history for eval datasets
bun run dev -- trends # score history and trend analysis
bun run dev -- ci # run all harvested tasks, fail on regression
bun run dev -- watch # watch instruction files, re-lint on save
bun run dev -- update # self-update to the latest release
bun run build # compile to binary (current platform)
bun run build:linux # cross-compile linux-x64
bun run build:darwin # cross-compile darwin-arm64
bun run build:darwin-x64 # cross-compile darwin-x64
bun run lint # biome check
bun run typecheck # tsc --noEmit
bun run check # lint + typecheck + test (CI equivalent)
Architecture
src/
├── lint/ # Static analysis rules (Phase 1)
├── run/ # Eval runner, worktree, scorer (Phase 2)
├── harness/ # Adapter pattern: claude-code, cursor, opencode, windsurf, copilot, generic, mock
├── store/ # Result persistence + comparison (Phase 2)
├── harvest/ # Git history mining: detect AI commits, emit task YAML (Phase 3)
├── config/ # Zod schema, YAML loader
├── markdown/ # Parser, section extractor, frontmatter
├── output/ # Console, JSON, markdown formatters
├── commands/ # CLI command handlers (init, doctor, lint, run, results, compare, harvest, ci, trends, watch, update)
└── utils/ # Logger, glob, path helpers
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.
- 4d ago First seen · 94 lines · 1,086 tokens per session scan C 1a5a591e568c
agenteval CLAUDE.md is an instructions file published in the GitHub repository lukasmetzler/agenteval (6 stars, last pushed 2mo ago), licensed MIT. It adds 1,086 tokens to every session, about $0.0054 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
openbench AGENTS.md
Instructions for minghinmatthewlam/openbench, covering openbench — agent context, local execution context, what openbench is, execution ownership and product goals (the two things we are building toward).
ai-shakespii AGENTS.md
Instructions for ai-creed/ai-shakespii, covering agents, project summary, non-negotiable decisions, source of truth and repo layout.
Math-To-Manim tests.instructions.md
Guidelines for writing and maintaining tests in Math-To-Manim.
aws-bench AGENTS.md
AGENTS.md instructions for aws-bench/aws-bench, covering agents.md, setup, commands, project structure and code style.
agentconfig.org AGENTS.md
Instructions for agentconfig/agentconfig.org, covering agent instructions for agentconfig.org, project overview, target audience, site structure and tech stack.
sandy CLAUDE.md
Instructions for jamestelfer/sandy, covering claude.md, voice, what is sandy, project layout and dev commands.