name: _convert-skill-from-codex description: > Convert a Codex skill to Claude Code format. You are Claude Code receiving a Codex skill and rewriting it for yourself. Use when the user wants to port a Codex skill to Claude Code.
Convert Codex Skill to Claude Code
You are Claude Code. You're receiving a Codex skill and need to rewrite it for yourself.
Finding Codex Skills
Codex skills are located at ~/.codex/skills/. Each skill is a directory containing:
SKILL.md- The main skill definition- Additional files - Templates, scripts, examples, or other supporting files
To list available Codex skills:
ls ~/.codex/skills/
Initial Response
If the user has not specified which skill to convert, list the available Codex skills and ask them to choose:
I'll help you convert a Codex skill to Claude Code format.
Available Codex skills:
[list from ~/.codex/skills/]
Which skill would you like me to convert?
Then wait for the user's input before proceeding.
Your Approach
Don't think about "adding Claude features to a Codex skill." Instead:
- Read the Codex skill to understand what it accomplishes
- Discover your available capabilities (sub-agents, tools)
- Write a fresh Claude Code skill that achieves the same goal using your native capabilities
- Express the workflow in your idiom
Process
Step 1: Understand the Source Skill
Read the Codex skill from ~/.codex/skills/<skill-name>/. Start with SKILL.md:
- What is this skill trying to help the user accomplish?
- What are the key steps in the workflow?
- What constraints or guidelines matter?
Focus on the INTENT, not the Codex-specific implementation details.
Step 2: Discover Your Capabilities
Discover available sub-agents:
Check ~/.claude/agents/ for available agents you can use:
ls ~/.claude/agents/
Each agent file (.md) has YAML frontmatter describing:
name: The agent identifier (use this with the Task tool)description: What the agent specializes intools: What tools the agent has access to
Read the agent files to understand what specialized agents are available to you. Common patterns include agents for:
- Codebase exploration and file discovery
- Code analysis and understanding
- Pattern finding and examples
- Web research (if available)
Your core capabilities:
- Task tool: Spawn sub-tasks that run in parallel with their own context
- Background execution: Run long tasks in background while continuing work
- Tool restrictions: Give sub-tasks limited tool access (e.g., read-only)
- Direct tools: Grep, Glob, Read, Edit, Write, Bash, etc.
Step 3: Write Your Version
Write a skill that accomplishes the same intent using your capabilities.
Frontmatter:
---
name: personal:<skill-name>
description: >
[What the skill does - write naturally, don't copy Codex's description]
---
Workflow: Express the steps in your natural idiom:
- You can spawn sub-tasks via the Task tool for parallel or isolated work
- Reference any available agents you discovered in Step 2 that would help
- You can run tasks in the background
- You can restrict tools for sub-tasks (e.g., read-only analysis)
Consider using sub-tasks when:
- Multiple independent operations could run in parallel
- High-volume output would clutter your main context
- Operations benefit from restricted tool access
- Long-running tasks could run in background
- A specialized agent exists that's well-suited for the task
Keep it sequential when:
- Steps depend on each other's results
- The workflow is simple and linear
- Sub-tasks would add complexity without benefit
- No specialized agents would help
Step 4: Text Replacements
Apply these replacements where they appear:
| Codex | Claude Code |
|---|---|
Codex (the agent) |
Claude Code |
Codex attribution |
Claude attribution |
Generated with Codex |
Generated with Claude |
list-codex-sessions |
list-claude-sessions |
read-codex-session |
read-claude-session |
shell_command |
appropriate tool (Bash, Read, etc.) |
Step 5: Handle Supporting Files
For each additional file in the skill directory:
- Templates: Apply text replacements
- Scripts: Update any Codex-specific commands
- Examples: Update to reflect Claude Code patterns
Step 6: Write and Validate
Write all files to
~/.claude/skills/<skill-name>/Validate:
- No remaining "Codex" references (except when discussing cross-agent work)
- Workflow uses your native patterns naturally
- Any referenced agents actually exist in
~/.claude/agents/ - YAML frontmatter is valid
- All source files accounted for
Summarize:
Converted: [skill name] Intent: [what the skill accomplishes] Key adaptations: - [How you expressed the workflow in your idiom] - [Any sub-agents you leveraged and why] - [Text replacements made] Output: [path]
Example
Codex skill says:
Research the codebase:
- Use `rg` to find files
- Read relevant files with `shell_command cat`
- Keep notes as you go
If you have relevant agents, you might write:
Research the codebase:
- Use a codebase exploration agent (if available) to find relevant files
- Use an analysis agent (if available) to understand implementation details
- Synthesize findings from sub-tasks
If sub-tasks aren't warranted or no relevant agents exist:
Research the codebase:
- Use Grep and Glob to find relevant files
- Read the files to understand the implementation
- Keep notes with file:line references
Choose based on whether parallelization actually helps the workflow and what agents are available to you.