Agentic Harness Engineering

Skill by ara.so — AI Agent Skills collection.

Overview

Agentic Harness Engineering (AHE) is an observability-driven system for automatically evolving the harness around a coding agent. The base LLM model remains frozen while AHE iteratively improves the harness components: system prompts, tool descriptions, tool implementations, middleware, skills, sub-agents, and long-term memory.

AHE operates through a three-phase loop:

1. Evaluate — Run the agent over a dataset, capture full traces
2. Analyze — Distill traces into root-cause reports using Agent Debugger
3. Improve — Evolve Agent proposes evidence-backed harness edits

Key achievements:

• Lifts GPT-5.4 from 69.7% → 77.0% pass@1 on Terminal-Bench 2 over 10 iterations
• Reaches 84.7% ± 2.1% with GPT-5.5 (ranked #3 on TB2 leaderboard)
• Produces transferable harnesses that work across models and benchmarks

Installation