skills.rest
ax-llm
ax-llmOFFICIAL
Explore all skills by this author →
GitHub

ax-agent-optimize

Generate Ax agent optimization code for agent.optimize workflows.

2.9k|186|Updated Feb 23, 2023
Search Tags:#evaluation#tool use#agent optimization#serialization#judge#ax
One-click install
npx skills add https://github.com/ax-llm/ax --skill ax-agent-optimize-ax-llm
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ax-agent-optimize
Source: https://github.com/ax-llm/ax/tree/main/website/static/typescript/.well-known/agent-skills/ax-agent-optimize
Command: npx skills add https://github.com/ax-llm/ax --skill ax-agent-optimize-ax-llm

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps generate correct Ax agent optimization code for improving tool use, response quality, delegation behavior, and reusable optimized artifacts without guessing the right workflow.

Core Features & Use Cases

  • Agent Optimization Setup: Builds practical agent.optimize workflows for actor or responder tuning.
  • Evaluation Design: Chooses between built-in judge scoring and deterministic metrics based on the task.
  • Artifact Save and Replay: Supports serializing optimized programs, restoring them later, and applying improvements safely.
  • Use Case: An engineer wants to improve an office assistant agent that sends emails, uses tools, and delegates sub-tasks, then save the tuned behavior for reuse.

Quick Start

Ask the skill to generate an agent.optimize example for your Ax agent, including the right evaluation strategy, task format, and optimized-program save and load steps.

Frequently Asked Questions about ax-agent-optimize

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

FAQPage Schema
How do I optimize an Ax agent for better tool use and response quality?▼

You optimize an Ax agent by generating an agent.optimize workflow that tunes tool-use and responder behavior using judge-based evaluation or deterministic metrics, then serializes the optimized program for reuse.

How do I set up judge-based evaluation for agent optimization?▼

Judge-based evaluation for agent optimization is set up by providing precise task records and judge options to the agent.optimize workflow, which scores responder outputs to guide tuning iterations.

Can I save and reload an optimized Ax agent program?▼

Yes, you can serialize optimized Ax agent programs to artifacts and restore them later, allowing you to safely apply previously tuned behaviors like tool use and delegation without re-running evaluations.

What's the best way to evaluate an agent that delegates sub-tasks?▼

The best way to evaluate delegation-heavy agents is using the agent.optimize workflow with eval-safe tools and precise task records, ensuring delegated sub-tasks are measured by judge scoring or deterministic metrics.

When should I use deterministic metrics instead of a judge for Ax agent evaluation?▼

You should use deterministic metrics instead of a judge when your agent optimization task has measurable, objective outcomes, switching to judge-based scoring when response quality requires subjective evaluation.

Related Skills

Explore More Software Engineering Skills→
jadeclijadecli
1

ax-agent-optimize

Generates AxAgent tuning and evaluation code for agent.optimize workflows.

Official
Advanced
project-axproject-ax
5

ax-cli

Execute AX CLI commands and manage npm provider packages.

Official
Advanced
ax-llmax-llm
2.9k

ax-java-agent-optimize

Configure Java Ax agent optimization runs with evaluators and bounded budgets.

Official
Intermediate
samibrahmisamibrahmi

security_scan

Run three security scanners in parallel and aggregate prioritized risk summaries.

Community
Advanced