Skip to content

CLAUDE.md

Latest commit

 

History

History
73 lines (56 loc) · 3.82 KB

CLAUDE.md

File metadata and controls

73 lines (56 loc) · 3.82 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Overview

This is the FlagOS Skills repository — a collection of reusable Agent Skills for FlagOS, a fully open-source AI system software stack for heterogeneous AI chips. Skills are folder-based capability packages that teach agents to reliably perform domain-specific tasks (model migration, deployment verification, complex workflows).

This repository follows the Agent Skills open standard, aligned with anthropics/skills, huggingface/skills, and openai/skills.

There is no build system or package manager. This is a pure content repository of Markdown-based skill definitions with supporting scripts and resources.

Repository Structure

├── .claude-plugin/          # Claude Code plugin manifest
│   └── marketplace.json     # Plugin bundles for Claude Code marketplace
├── .cursor-plugin/          # Cursor plugin manifest
│   ├── marketplace.json     # Cursor marketplace listing
│   └── plugin.json          # Plugin metadata
├── agents/                  # Codex / Gemini CLI fallback
│   └── AGENTS.md            # Skills index for agents without native support
├── assets/                  # Repository-level static resources
├── contributing.md          # Contribution guidelines
├── gemini-extension.json    # Gemini CLI extension manifest
├── scripts/                 # Repository-level utility scripts
│   └── validate_skills.py   # Batch validate all skills
├── skills/                  # Actual skill directories
│   ├── model-migrate-flagos/    # Example: model migration workflow
│   └── ...
├── spec/                    # Agent Skills standard & local conventions
│   ├── README.md            # Conventions and cross-reference to standards
│   └── agent-skills-spec.md # Redirect to agentskills.io/specification
└── template/                # Template for creating new skills
    └── SKILL.md

Skill Structure

Each skill is a directory under skills/ containing at minimum a SKILL.md file with:

  1. YAML frontmatter (required fields):
    • name: unique identifier, lowercase with hyphens (e.g., model-migrate-flagos)
    • description: one or two sentences explaining what the skill does and when to trigger it
  2. Optional frontmatter: license, compatibility, metadata, allowed-tools, user-invokable, argument-hint
  3. Markdown body: instructions for the agent — applicable scenarios, constraints, steps, and examples (at least 2–3)
  4. Optional subdirectories: references/, scripts/, assets/, examples/

Creating a New Skill

mkdir skills/<skill-name>
cp template/SKILL.md skills/<skill-name>/SKILL.md
# Edit the SKILL.md: update frontmatter and write agent instructions
# Validate: python scripts/validate_skills.py

See contributing.md for full guidelines.

Naming Conventions

  • Directory names and name fields: lowercase + hyphens (my-skill-name)
  • Max 64 characters for name
  • Avoid vague names like misc, tmp, v1
  • FlagOS skills use -flagos suffix: model-migrate-flagos, skill-creator-flagos

Key Conventions

  • Additional scripts/resources can be placed alongside SKILL.md but must be documented in the markdown body
  • When skills conflict, the more specific/business-contextual skill takes priority
  • See spec/README.md for detailed conventions and the Agent Skills specification
  • Plugin bundles are defined in .claude-plugin/marketplace.json and .cursor-plugin/
  • Run python scripts/validate_skills.py before submitting PRs