agent-lightning

Automate AI agent training setup and optimization with the Agent Lightning framework.

264|11|Updated Apr 25, 2026
One-click install
npx skills add https://github.com/rkz91/coco --skill agent-lightning
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-lightning
Source: https://github.com/rkz91/coco/tree/main/skills/agent-lightning
Command: npx skills add https://github.com/rkz91/coco --skill agent-lightning

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Train and optimize AI agents using the Agent Lightning framework to streamline training, tracing, resource management, and reward design across multi-agent pipelines.

Core Features & Use Cases

  • Architecture Flow and central tracing with LightningStore for scalable agent training.
  • Instrumentation of agent code with emit_xxx calls for observability and reward signals.
  • Support for Reinforcement Learning (GRPO/PPO) and Automatic Prompt Optimization (APO) workflows, plus framework adapters.
  • End-to-end trainer configuration and customization with configurable runners, algorithms, and stores.

Quick Start

Install the package, configure a minimal trainer with a simple store, and run a basic training loop.

Frequently Asked Questions about agent-lightning

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

FAQPage Schema
How do I set up multi-agent pipelines for AI training and tracing?

To set up multi-agent pipelines for AI training and tracing, configure a trainer with LightningStore as the central tracing hub, then execute a basic training loop to automate scalable agent training and resource management.

What is the best way to instrument agent code for observability and reward signals?

The best way to instrument agent code for observability and reward signals is to integrate emit_xxx calls directly into your pipeline, enabling detailed logging and evaluation for reinforcement learning workflows.

Can I use GRPO and PPO algorithms for automatic prompt optimization in multi-agent systems?

Yes, you can use GRPO and PPO algorithms for automatic prompt optimization in multi-agent systems by configuring the framework adapters to support both reinforcement learning and APO workflows.

How does LightningStore support scalable agent training and instrumentation?

LightningStore supports scalable agent training by acting as a central tracing store for architecture flow, capturing instrumentation data from emit_xxx calls to ensure comprehensive observability and reward signal tracking.

Do I need specific framework adapters to run reinforcement learning and APO workflows?

Yes, you need modular framework adapters to run reinforcement learning and APO workflows, as these adapters ensure proper integration with configurable runners, algorithms, and stores for end-to-end trainer customization.