eval-harness

eval-harness is a skill for Claude Code from Jarroslav/agentic-os. It costs 91 tokens per session (2,096 once invoked), scanned A, original, Apache-2.0.

A framework for testing coding-agent skills and instruction files with two layers: automatic structure checks and optional tests judged by another language model.

In plain words
What is it for?
Use it to add regression checks, test skill or agent instructions, run cheap contract checks in commits, and evaluate behavior with repeated judge cases.
Why use it?
It catches both broken files and behavior changes that leave the files present but make the skill work incorrectly.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: mentions Claude Code.

Part of the agentic-qe plugin — 2 skills 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 skills/jarroslav/agentic-os/eval-harness
Any agent
npx skills add Jarroslav/agentic-os --skill eval-harness
Clone the repo
git clone --depth 1 https://github.com/Jarroslav/agentic-os

Made for: Claude Code.

Or install agentic-qe, the plugin that ships this one along with the rest of its 2 skills.

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-harness

README.md
[![agentmods](https://agentmods.dev/badge/skills/jarroslav/agentic-os/eval-harness.svg)](https://agentmods.dev/skills/jarroslav/agentic-os/eval-harness)
Your own site
<a href="https://agentmods.dev/skills/jarroslav/agentic-os/eval-harness"><img src="https://agentmods.dev/badge/skills/jarroslav/agentic-os/eval-harness.svg" alt="Measured on agentmods" height="20"></a>
Per session 91 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,096 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 5.1 $0.00091 $0.02096
Opus 5 $0.00046 $0.01048
Sonnet 5 $0.00018 $0.00419
Haiku 4.5 $0.00009 $0.00210

Measured 5d ago against content hash 0d58f1a89cec, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

eval-harness 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.

The scan reads SKILL.md. This mod also ships 6 executable files (assets/python/llm_eval_runner.py, assets/python/runner.py, assets/python/test_runner.py, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

plugins/agentic-qe/skills/eval-harness/SKILL.md · 128 lines

How it starts

The opening of the file, as written. The whole thing — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Eval Harness

Bootstrap repo-wide evaluation for skill and agent instruction files. Two layers, one spec file per target.

Layer Mechanism Failure class caught Cost profile
Contract Code-only checks, zero LLM calls Structural rot: missing files, edited-away instructions, oversized bodies, broken helper scripts Cheap — safe to run on every commit
Judge Candidate model executes the skill against a case prompt; judge model grades output vs assertions Behavioral drift: skill still loads but no longer behaves as designed Model calls — on demand or scheduled

Both layers exist because they detect different failure classes. Route structure to code checks and behavior to the judge: never regex-grade prose, never burn a model call to confirm a file exists.

Both layers read the same spec, eval/evals.json, colocated inside each target's directory. A spec with contracts and no behavioral cases is valid. Each judge case runs N repetitions, so flakiness surfaces as a pass rate instead of a binary verdict.

eval/ here is not evals/ in the agentic-os marketplace repo, and the near-collision is deliberate to name rather than rename. This skill writes <target>/eval/evals.json into your repo. The marketplace repo's own CI (tests/lib/check-skill-contract.py) separately requires <skill>/evals/evals.json for the skills it ships. Different scope, different owner, no conflict — except when you point this skill at the marketplace repo itself, where you will end up with both. In that one case keep the plugin's evals/ as the shipped contract and treat the generated eval/ as local scaffolding.

Blast radius: R2 — this skill writes harness code, spec files, and config into the repo. Judge runs place outbound model calls; automation wiring is opt-in and confirmed first. It never creates real credential files and never gates merges on paid model runs.

Inputs

Collect during the interview phase. Ask only what the repo has not already answered; confirm inferences rather than re-asking.

Read the full file on GitHub · 128 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 · 128 lines · 0 tokens per session scan A 0d58f1a89cec

Subscribe to this mod's changes

eval-harness is a skill published in the GitHub repository Jarroslav/agentic-os (0 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 91 tokens to every session and 2,096 once invoked, about $0.0005 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 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

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

chronicle

Analyze Copilot session history for standup reports, usage tips, session search, and session reindexing. Use when the user asks for a standup, daily summary, usage tips, workflow recommendations, wants to search or find past sessions by keyword/file/PR, wants to reindex their session store, or asks about deleting…

microsoft/vscode · 72 tokens