Supported Agents for Garden Skills: Claude Code, Cursor, Codex, and Generic Agents

Garden Skills supports Claude Code, Cursor, Codex, and any generic agent implementing the SKILL.md specification, with each platform reading structured prompts from dedicated directories like .claude/skills/, .agents/skills/, or .codex/skills/.

The ConardLi/garden-skills repository maintains a curated collection of production-ready agent skills using the standardized SKILL.md format. Understanding the supported agents for garden skills allows developers to deploy the same structured system prompts across multiple AI coding tools without modification.

Supported AI Agents and Installation Paths

According to the repository's README.md (lines 5 and 419-449), Garden Skills explicitly targets the following agents with designated installation folders:

Agent Installation Directory Status
Claude Code .claude/skills/<skill-name>/ ✅ Tested
Cursor .agents/skills/<skill-name>/ ✅ Tested
Codex .codex/skills/<skill-name>/ ✅ Tested
Generic Agents .agents/skills/<skill-name>/ ✅ Supported

Each agent reads the SKILL.md file contained within these directories to load the skill's structured system prompt.

How Each Agent Consumes SKILL.md Files

Garden Skills operates through a file-based architecture where agents consume plain-text SKILL.md files from specific paths. Because skills contain no runtime code, the same files work across all supported platforms.

Claude Code Integration

Claude Code (Anthropic) searches for skill definitions inside the .claude/skills/ directory. When you place a skill folder containing a SKILL.md file at this path, Claude Code automatically loads the structured prompt and applies it to generate or refactor code.

Cursor Integration

Cursor reads skill definitions from .agents/skills/. By unpacking a skill package into this directory, Cursor recognizes the SKILL.md format and loads the prompt for subsequent coding tasks. This path also serves as the generic location for any agent supporting the Agent Skills specification.

Codex Integration

Codex-compatible tools (OpenAI) look for skills in .codex/skills/. The repository confirms this path is tested and functional for Codex deployments, allowing the same skill files to power OpenAI's agentic coding tools.

Generic Agent Support

Any agent implementing the public Agent Skills specification (agentskills.io) can consume skills from the .agents/skills/ path. Since the format is pure text with no execution dependencies, custom agents and third-party tools can parse the SKILL.md files directly.

Practical Installation Examples

The following commands demonstrate how to install the same skill (e.g., web-design-engineer) for different agents using the paths defined in README.md:

Installing for Claude Code


# Download the skill package

wget https://github.com/ConardLi/garden-skills/releases/download/v1.0/web-design-engineer.zip

# Extract to Claude Code's designated directory

unzip -q web-design-engineer.zip -d .claude/skills/

After extraction, Claude Code reads the skill from .claude/skills/web-design-engineer/SKILL.md.

Installing for Cursor


# Extract to the generic agents directory used by Cursor

unzip -q web-design-engineer.zip -d .agents/skills/

Cursor automatically detects the skill at .agents/skills/web-design-engineer/ and loads the prompt on startup.

Installing for Codex


# Extract to Codex's specific skill directory

unzip -q web-design-engineer.zip -d .codex/skills/

Codex-compatible tools load the skill from .codex/skills/web-design-engineer/SKILL.md.

Key Repository Files

The following files in ConardLi/garden-skills define and demonstrate agent support:

  • README.md — Contains the agent compatibility matrix and installation instructions for Claude Code, Cursor, and Codex (referenced at lines 5, 419, and 449).
  • skills/web-design-engineer/README.md — Example skill demonstrating concrete usage across all three major agents.
  • skills/kb-retriever/README.md — Shows generic agent compatibility using the .agents/skills/ path.
  • package.json — Describes the repository as a collection of Agent Skills targeting the supported platforms.

Summary

  • Garden Skills officially supports Claude Code, Cursor, Codex, and generic SKILL.md-compatible agents as documented in the repository's README.md.
  • Each platform uses a specific installation directory: .claude/skills/ for Claude Code, .codex/skills/ for Codex, and .agents/skills/ for Cursor and generic agents.
  • Skills utilize the standardized SKILL.md format, enabling cross-agent compatibility without code modifications or runtime dependencies.
  • The repository provides tested examples in skills/web-design-engineer/ and other directories demonstrating multi-agent deployment.

Frequently Asked Questions

What AI agents are officially supported by Garden Skills?

Garden Skills officially supports Claude Code (Anthropic), Cursor, and Codex (OpenAI), along with any generic agent implementing the Agent Skills specification. According to the README.md in the ConardLi/garden-skills repository, all three major agents have tested integration paths using their respective .claude/skills/, .agents/skills/, and .codex/skills/ directories.

Where should I install Garden Skills for Cursor versus Claude Code?

For Cursor, install skills to .agents/skills/<skill-name>/ as documented in the repository's compatibility table. For Claude Code, use .claude/skills/<skill-name>/ instead. Both platforms read the same SKILL.md file from these distinct paths, allowing you to maintain parallel installations if you use multiple agents on the same project.

Can the same skill files be used across different agents without modification?

Yes. Because Garden Skills relies on the standardized SKILL.md format—a plain-text specification without runtime code—the same skill folder can be copied to .claude/skills/, .codex/skills/, or .agents/skills/ without any changes. The repository's skills/ directory contains examples like web-design-engineer that function identically across all supported platforms.

Does Garden Skills require any runtime execution or dependencies?

No. Garden Skills consist entirely of structured text files following the SKILL.md specification. Agents consume these files as system prompts, meaning there is no executable code, runtime dependencies, or build steps required. This architectural choice ensures universal compatibility across Claude Code, Cursor, Codex, and future agents adopting the standard.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →