📘
AI GO! skills give your AI coding agent the knowledge it needs to work with the AI Platform.

The lf skills command provides a skill definition and a curated list of reference materials that teach your agent how to use the lf CLI, write YAML specs, and run evaluations.

What does the agent learn?

The skill context covers:

Core concepts: how to use AI apps as projects, how to define and run evaluations, how the entities relate to each other.
CLI usage: how to use all lf commands and flags.
YAML syntax: how to use $ref, !include, $provider, environment variables, and templates.
Entity specs: complete spec for every entity type.

In addition to the skill, the agent also has access to two reference resources:

AI Atlas: A library of curated evaluation templates covering safety, robustness, performance, and more. Each template includes a ready-to-use run.yaml, evaluation.yaml, and supporting specs. The agent uses these as starting points for defining new evaluations.
AI GO! Integrations: Real-world YAML examples organized into guides, tutorials, and references. The agent uses these to learn the correct YAML syntax by example.

Both commands below download the skill definition and reference resources, cached per SDK version. Re-running either command downloads fresh copies.

Install into your project — writes a SKILL.md file so your agent discovers the skill automatically.
Load on the fly — prints the skill context to standard output without writing files into your project.

Install into Your Project

Run lf skills install with your agent type. For supported agents beyond Claude Code and OpenCode, see the CLI reference.

lf skills install --agent <agent>

Load on the Fly

Run lf skills to output the skill context directly.

lf skills

To use this approach, prompt your agent:

Please run lf skills to load AI GO! evaluation skills.