meta-agent-harness-optimizer

Community

Optimize harnesses from traces and prompts.

AuthorAskTinNguyen
Version1.0.0
Installs0

System Documentation

What problem does it solve?

Continual harness optimization for multi-agent systems by enabling a reusable pattern to study, port, and operationalize harness designs from traces, prompts, and tool surfaces.

Core Features & Use Cases

  • Provides a structured loop pattern where a proposer writes harness config variants, which are validated and evaluated against benchmarks.
  • Captures and reason over artifact memory (traces, scores, summaries) to guide improvements and ensure auditability.
  • Enables porting and extension of the canvas-org/meta-agent architecture into environments like Vesper, with filesystem-backed experience stores and modular components.

Quick Start

Run the outer loop against a benchmark to begin evolving and evaluating harness configurations.

Dependency Matrix

Required Modules

None required

Components

Standard package

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: meta-agent-harness-optimizer
Download link: https://github.com/AskTinNguyen/vesper-team-skills/archive/main.zip#meta-agent-harness-optimizer

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 471,000+ vetted skills library on demand.