reference

Document HarnessX component authoring shapes and workflows for tools, processors, and prompts.

415|49|Updated Apr 30, 2026
One-click install
npx skills add https://github.com/Darwin-Agent/HarnessX --skill reference-darwin-agent
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: reference
Source: https://github.com/Darwin-Agent/HarnessX/tree/main/harnessx/meta_harness/workspace/skills/reference
Command: npx skills add https://github.com/Darwin-Agent/HarnessX --skill reference-darwin-agent

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Provides the official guidelines for authoring HarnessX components (tools, processors, prompts) and for managing evolve rounds, ensuring consistency and reproducibility across benchmarks.

Core Features & Use Cases

  • Covers the @tool signature and TOOL_SPEC workflow to define tools clearly.
  • Documents MultiHookProcessor usage, hook dispatch tables, event fields, and messages-mutation contracts.
  • Explains system-prompt editing across Cases A/B/C, and how to align config.yaml shapes with the harness.
  • Serves as a hands-on reference when implementing lever types: Action, Control, Instruction, or Configuration.

Quick Start

Read this reference when starting a new lever or updating an existing component to align with HarnessX conventions.

Frequently Asked Questions about reference

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

FAQPage Schema
How do I define tools for HarnessX component authoring?

To define tools for HarnessX component authoring, use the @tool signature and follow the TOOL_SPEC workflow to ensure tools are clearly defined and consistently structured across benchmarks.

What is the MultiHookProcessor messages-mutation contract?

The MultiHookProcessor messages-mutation contract defines how event fields and hook dispatch tables mutate messages during processing, ensuring deterministic behavior across harness integrations and evolve rounds.

How does system-prompt editing work across Cases A, B, and C?

System-prompt editing across Cases A, B, and C involves aligning config.yaml shapes with the harness to support deterministic evolve rounds, specifying exactly how prompt modifications are applied in each scenario.

Do I need a config.yaml file to organize HarnessX component configuration?

Yes, you need a config.yaml file to organize HarnessX component configuration, as it specifies dependencies, available components, and safety considerations required to enable secure and reproducible integration.

What are the lever types available when implementing HarnessX components?

The lever types available when implementing HarnessX components are Action, Control, Instruction, and Configuration, each serving distinct roles in managing workflows and system behavior.

When should I use official guidelines for HarnessX component authoring?

Use official guidelines for HarnessX component authoring when starting a new lever or updating an existing component to align with HarnessX conventions and ensure reproducibility across benchmarks.