Objective
Make this configuration repository work with both Claude Code and OpenAI Codex, so users can apply the same agents, skills, rules, and behavioral constraints regardless of which AI coding assistant they use.
Background
Currently, this repository is Claude Code-specific. The agents, skills, and rules are already tool-agnostic in concept, but the installation, configuration files, and documentation assume Claude Code only.
Required Changes
1. Extract Generic Core Instructions
What: Create a new generic instruction file that contains tool-agnostic behavioral constraints.
Why: The current CLAUDE.md mixes tool-agnostic principles with Claude Code-specific references. We need a shared source of truth.
Changes:
- Extract the core behavioral constraints (think before coding, simplicity first, surgical changes, goal-driven execution) into a new generic file
- Keep the Claude Code-specific file as a lightweight wrapper that references the generic core
- Both tools can then reference the same core principles
2. Separate Tool-Specific Configuration Files
What: Clearly distinguish between Claude Code config and future Codex config.
Why: Each tool has its own settings format and location. Currently, settings and keybindings are unlabeled, making it unclear they're Claude Code-specific.
Changes:
- Rename existing configuration files to indicate they're for Claude Code
- Create a dedicated directory for OpenAI Codex-specific configuration
- Document what each configuration file controls
3. Create Installation Scripts
What: Provide automated installation for each tool instead of manual copy commands.
Why: Users shouldn't need to remember the correct paths and files for each tool. Scripts reduce errors and improve user experience.
Changes:
- Create a Claude Code installation script that copies files to the correct home directory location
- Create an OpenAI Codex installation script that copies files to Codex's configuration location
- Include platform detection for macOS, Linux, and optionally Windows
- Add verification steps to confirm successful installation
4. Add Codex-Specific Adapter Files
What: Create configuration files in OpenAI Codex's expected format.
Why: Codex likely has different configuration syntax and requirements than Claude Code.
Changes:
- Research Codex's configuration format and directory structure
- Create Codex configuration files that reference the shared core instructions
- Ensure agents and skills are accessible in Codex's format
- Document any feature limitations or differences
5. Update Documentation
What: Revise README to present this as a multi-tool repository.
Why: Users need to understand they can choose which tool(s) to configure.
Changes:
- Update the introduction to mention both supported tools
- Provide separate installation instructions for each tool
- Document verification steps for each tool
- Explain that the same agents, skills, and rules work across both tools
- Note any feature differences or limitations between tools
Success Criteria
Research Needed
Before implementation, we need to determine:
- Codex configuration location — Where does OpenAI Codex look for configuration files?
- Codex format support — Does Codex support agent definitions and slash commands like Claude Code does?
- Codex settings format — What does a Codex settings file look like?
- Feature parity — Are there features that work in Claude Code but not in Codex (or vice versa)?
- Installation paths — Do these vary by platform (macOS vs Linux vs Windows)?
Out of Scope
- Support for other AI coding tools (Cursor, Copilot, Aider, etc.) — this can be addressed in future issues if desired
- Rewriting existing agents, skills, or rules — they're already tool-agnostic
- Breaking changes to the repository structure beyond what's necessary
Implementation Approach
This can be done incrementally:
- Phase 1: Create generic core instructions and update Claude Code setup (backward compatible)
- Phase 2: Research and add OpenAI Codex support
- Phase 3: Update documentation and add installation scripts
- Phase 4: Test on multiple platforms and gather feedback
Objective
Make this configuration repository work with both Claude Code and OpenAI Codex, so users can apply the same agents, skills, rules, and behavioral constraints regardless of which AI coding assistant they use.
Background
Currently, this repository is Claude Code-specific. The agents, skills, and rules are already tool-agnostic in concept, but the installation, configuration files, and documentation assume Claude Code only.
Required Changes
1. Extract Generic Core Instructions
What: Create a new generic instruction file that contains tool-agnostic behavioral constraints.
Why: The current
CLAUDE.mdmixes tool-agnostic principles with Claude Code-specific references. We need a shared source of truth.Changes:
2. Separate Tool-Specific Configuration Files
What: Clearly distinguish between Claude Code config and future Codex config.
Why: Each tool has its own settings format and location. Currently, settings and keybindings are unlabeled, making it unclear they're Claude Code-specific.
Changes:
3. Create Installation Scripts
What: Provide automated installation for each tool instead of manual copy commands.
Why: Users shouldn't need to remember the correct paths and files for each tool. Scripts reduce errors and improve user experience.
Changes:
4. Add Codex-Specific Adapter Files
What: Create configuration files in OpenAI Codex's expected format.
Why: Codex likely has different configuration syntax and requirements than Claude Code.
Changes:
5. Update Documentation
What: Revise README to present this as a multi-tool repository.
Why: Users need to understand they can choose which tool(s) to configure.
Changes:
Success Criteria
Research Needed
Before implementation, we need to determine:
Out of Scope
Implementation Approach
This can be done incrementally: