skills.rest
tmuskal
tmuskalCOMMUNITY
Explore all skills by this author →
𝕏@tmuskalGitHub

arc-cross-harness

Generate cross-harness benchmarking instructions for ARC-AGI and compare results.

3|1|Updated Apr 11, 2026
Search Tags:#codex#benchmarking#gemini#opencode#arc-agi
One-click install
npx skills add https://github.com/tmuskal/arc-agi-benchmarker --skill arc-cross-harness
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: arc-cross-harness
Source: https://github.com/tmuskal/arc-agi-benchmarker/tree/main/plugins/arc-agi-benchmarker/skills/cross-harness
Command: npx skills add https://github.com/tmuskal/arc-agi-benchmarker --skill arc-cross-harness

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires arc-agi, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill allows users to benchmark ARC-AGI across different harnesses like Codex, Gemini CLI, and OpenCode, enabling them to compare performance and results.

Core Features & Use Cases

  • Cross-Harness Benchmarking: Generate benchmark instructions for various harnesses and import results.
  • Instruction Generation: Create instruction documents and helper scripts for harness-specific benchmarking.
  • Result Comparison: Compare benchmark results across different harnesses.
  • Use Case: A researcher might use this Skill to compare the performance of ARC-AGI on Codex and OpenCode, adjusting parameters to optimize performance.

Quick Start

To generate instructions for Codex benchmarking, run:

/arc-cross-harness generate codex

Frequently Asked Questions about arc-cross-harness

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I benchmark ARC-AGI performance across different coding harnesses?▼

To benchmark ARC-AGI across different harnesses, you can generate specific instruction documents and helper scripts for platforms like Codex, Gemini CLI, and OpenCode, then import and compare their results.

What's the best way to compare ARC-AGI results between Codex and OpenCode?▼

Comparing ARC-AGI results between Codex and OpenCode requires generating harness-specific benchmark instructions, running the tests, and importing the results to evaluate relative performance metrics.

Do I need a Python environment and the arc-agi package to run cross-harness benchmarking?▼

Yes, cross-harness benchmarking requires a Python environment with the arc-agi package installed to generate instructions, execute helper scripts, and import performance results for comparison.

Can I generate benchmarking instructions for Gemini CLI to test ARC-AGI?▼

Yes, you can generate benchmarking instructions and helper scripts specifically for Gemini CLI to test ARC-AGI by executing a simple generate command targeting the harness.

Why use cross-harness benchmarking instead of testing ARC-AGI on a single platform?▼

Cross-harness benchmarking identifies performance differences across Codex, Gemini CLI, and OpenCode, allowing researchers to adjust parameters and optimize ARC-AGI results for specific environments.

What does the arc-cross-harness Skill do for ARC-AGI evaluation?▼

The arc-cross-harness Skill facilitates ARC-AGI evaluation by generating benchmark instructions for Codex, Gemini CLI, and OpenCode, then importing and comparing the performance results.

Related Skills

Explore More Software Engineering Skills→
tmuskaltmuskal
3

arc-agi-benchmarker-setup

Install Python, create a virtual environment, and verify the ARC-AGI package setup.

Community
Advanced
tmuskaltmuskal
3

cross-harness

Automate LongMemEval benchmarking across Codex, Gemini, and OpenCode harnesses.

Community
Intermediate
mpazarynampazaryna
1

fork

Fork a new terminal session to execute commands or agentic coding tools.

Community
Intermediate
agentic-inagentic-in
577

LM Evaluation Harness

Benchmark LLM performance across 60+ academic benchmarks with reproducible scoring.

Official
Advanced