Skip to content

docs: add SKILL.md for Claude Code agent integration#433

Open
DavyMorgan wants to merge 5 commits intoalgorithmicsuperintelligence:mainfrom
DavyMorgan:docs/add-skill-md
Open

docs: add SKILL.md for Claude Code agent integration#433
DavyMorgan wants to merge 5 commits intoalgorithmicsuperintelligence:mainfrom
DavyMorgan:docs/add-skill-md

Conversation

@DavyMorgan
Copy link
Contributor

Summary

  • Add SKILL.md at the repo root — a Claude Code skill file with YAML frontmatter (name, description) and comprehensive agent instructions for using OpenEvolve
  • Covers the full workflow: scaffolding experiments, writing evaluators, configuring evolution, running jobs, inspecting checkpoints, resuming from checkpoints, and retrieving best evolved programs
  • Includes troubleshooting table, cost management tips, provider configuration examples, and references to all built-in examples

What is a Claude Code Skill File?

A skill file is a markdown document with YAML frontmatter that Claude Code automatically discovers and uses when relevant. When a user asks Claude Code to "evolve a sorting algorithm" or "optimize a GPU kernel with OpenEvolve," the skill file provides the agent with precise, up-to-date instructions — eliminating guesswork and hallucination.

This repo already has CLAUDE.md (developer guide for working on OpenEvolve) and .github/workflows/claude.yml (GitHub Action). The SKILL.md complements these by teaching agents how to use OpenEvolve as an end-user tool.

Why at root level?

The repo already places CLAUDE.md at root. SKILL.md follows the same convention for visibility. If the maintainers prefer .claude/skills/openevolve.md (the Claude Code directory convention), happy to move it there.

Test plan

  • Verify SKILL.md renders correctly on GitHub (YAML frontmatter displayed as metadata)
  • Confirm Claude Code discovers the skill when cloning the repo and asking about evolutionary optimization
  • Check that no existing CI workflows are affected (docs-only change)

🤖 Generated with Claude Code

Add a Claude Code skill file that provides comprehensive agent
instructions for using OpenEvolve. This makes the tool ready-to-use
for AI coding agents by documenting the full workflow: scaffolding
experiments, writing evaluators, configuring evolution, running jobs,
inspecting checkpoints, and retrieving evolved programs.

The skill file uses YAML frontmatter (name, description) so that
Claude Code can automatically discover and activate it when users
ask about evolutionary code optimization.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Copilot AI review requested due to automatic review settings March 7, 2026 04:51
Copy link

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Adds a Claude Code “skill file” (SKILL.md) intended to teach agents how to use OpenEvolve as an end-user tool (scaffolding experiments, writing evaluators, configuring/running evolution, resuming from checkpoints, and inspecting results).

Changes:

  • Add root-level SKILL.md with YAML frontmatter for Claude Code discovery.
  • Document end-to-end OpenEvolve workflow, including CLI/API usage, configuration patterns, troubleshooting, and built-in examples.

You can also share your feedback on Copilot code review. Take the survey.

DavyMorgan and others added 4 commits March 7, 2026 00:06
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
- Fix core concepts: config.yaml is optional, not required
- Clarify EVOLVE-BLOCK markers are LLM guidance, not enforced by the tool
- Fix fitness description: excludes MAP-Elites feature dimensions
- Remove non-numeric error field from evaluator example metrics dict
- Fix checkpoint structure: programs are .json, add best_program.py and
  best_program_info.json, add artifacts/ directory
- Fix output paths to be consistent (my_experiment/output throughout)
- Point to best_program.py instead of programs/<id>.py for result retrieval
- Remove plotly from visualizer install (Flask only, no plotly dependency)
- Clarify artifact config: evaluator.enable_artifacts + prompt.include_artifacts

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@codelion codelion added the enhancement New feature or request label Mar 7, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants