harness-design

Define tool space, observation format, and evaluation signals for LLM agent harnesses.

Updated Apr 3, 2026

Search Tags:#diagnosis #harness-design #llm-agent #tool-boundaries #observation-format #evaluation-signal

Ready to run

npx skills add https://github.com/Victoriakaey/build-reliable-agents --skill harness-design

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: harness-design
Source: https://github.com/Victoriakaey/build-reliable-agents/tree/main/skills/harness-design
Command: npx skills add https://github.com/Victoriakaey/build-reliable-agents --skill harness-design

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Harness design solves the common agent harness failures by clarifying the tool space, observation scope, and evaluation signals before prompts.

Core Features & Use Cases

Five-layer harness design: Action Space, Tool Boundaries, Observation Format, Evaluation Signals, and Harness Iteration.
Clear contracts and diagnostics for diagnosing misbehavior and guiding iterative improvements.
Use cases include designing a new agent harness, debugging existing agents, and hardening production-grade agent loops.

Quick Start

Outline the five harness layers for your current task and document the inputs, outputs, and success metrics for each layer.

Frequently Asked Questions about harness-design

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

FAQPage Schema

How do I design a reliable agent harness for LLM agents?▼

A reliable agent harness defines the tool space, observation format, and evaluation signals before writing prompts. You structure it across five layers—action space, tool boundaries, observation, evaluation, and iteration—to enforce strict input/output contracts and enable repeatable improvements.

Why does my LLM agent misbehave when calling tools?▼

LLM agent misbehavior often stems from unclear tool boundaries and observation scope. By applying strict per-action logging, error handling, and clear input/output contracts, you can diagnose exactly where the agent loop fails and guide iterative improvements.

What is the best way to structure evaluation signals for agent loops?▼

The best way to structure evaluation signals is to define them as a dedicated harness layer alongside action space and observation format. This ensures you capture specific success metrics for each layer, enabling repeatable diagnostics and feedback mechanisms during iteration.

How do I debug an existing LLM agent harness that keeps failing?▼

To debug a failing LLM agent harness, outline its five layers—action space, tool boundaries, observation, evaluation, and iteration—and document the inputs, outputs, and success metrics for each. This pinpoints contract violations and guides targeted fixes.

Do I need strict input/output contracts for production-grade agent loops?▼

Yes, strict input/output contracts are essential for production-grade agent loops. They ensure reliable action execution, enable accurate per-action logging, and provide the diagnostics needed to maintain stability across tool space and data flow.

When should I not use a five-layer harness design for my agent?▼

A five-layer harness design may be unnecessary for simple, single-tool tasks with minimal data flow. It is specifically built for designing new harnesses, debugging complex agents, and hardening production-grade loops involving multiple tools and feedback mechanisms.