1
schema: 1
2
name: eval-harness
3
owner: openagents
4
version: 1.0.0
5
kind: harness
6
title: Agent Eval Harness
7
summary: "Build a real eval suite for an agent or prompt: cases, graders, a baseline, and a regression gate"
8
license: MIT
9
tags: [evals, testing, harness, regression, quality]
10
runtimes: [claude-code, codex, openai-agents, langgraph, generic]
11
pricing:
12
  model: free
13
  amount_cents: 0
14
  currency: usd
15
entry: HARNESS.md
16
files:
17
  - HARNESS.md
18
  - grading.md
19
  - templates/cases.jsonl
20
inputs:
21
  - name: target
22
    type: string
23
    required: true
24
    description: "What is being evaluated: a prompt file, an agent entry point, or a graph node"
25
  - name: n_cases
26
    type: number
27
    required: false
28
    default: 30
29
    description: How many cases to build before measuring a baseline
30
requires: []
31