Category
Platform Support - Add support for Udio, new distributors, etc.
Which skill? (if enhancing existing)
None
Problem or use case
The project is currently tightly integrated with Claude Code, but much of the core — skills/, MCP server, references, templates, and music tooling — is already largely provider-neutral.
It would be useful to make these assets reusable from other Agent Skills runtimes, starting with OpenAI Codex, without regressing or replacing the existing Claude Code experience.
Proposed solution
Gradually separate provider-neutral functionality from Claude-specific runtime assumptions while keeping Claude Code first-class.
The intended architecture would be roughly:
- one shared
skills/ and MCP implementation
- existing Claude Code integration remains unchanged
- thin runtime-specific adapters/packaging for Codex or other compatible agents
- runtime-specific orchestration only where necessary
I would suggest implementing this incrementally rather than as one large port.
I'm willing to maintain and test the migration in a fork first, then submit focused PRs to develop if this direction is acceptable.
Alternatives considered
Maintain a separate Codex-specific fork/project.
That would work, but would duplicate a large amount of fast-moving skill and workflow content and create long-term synchronization overhead.
Additional context
#438 / #439 already established a small precedent for improving compatibility with another Agent Skills runtime (GitHub Copilot CLI) while preserving Claude Code behavior.
Before going further, I'd mainly like to know:
- Is optional support for additional Agent Skills runtimes an acceptable direction for this repository?
- If so, would you prefer that support to live here behind thin runtime adapters, or remain in a separate downstream project?
Contribution
Category
Platform Support - Add support for Udio, new distributors, etc.
Which skill? (if enhancing existing)
None
Problem or use case
The project is currently tightly integrated with Claude Code, but much of the core —
skills/, MCP server, references, templates, and music tooling — is already largely provider-neutral.It would be useful to make these assets reusable from other Agent Skills runtimes, starting with OpenAI Codex, without regressing or replacing the existing Claude Code experience.
Proposed solution
Gradually separate provider-neutral functionality from Claude-specific runtime assumptions while keeping Claude Code first-class.
The intended architecture would be roughly:
skills/and MCP implementationI would suggest implementing this incrementally rather than as one large port.
I'm willing to maintain and test the migration in a fork first, then submit focused PRs to
developif this direction is acceptable.Alternatives considered
Maintain a separate Codex-specific fork/project.
That would work, but would duplicate a large amount of fast-moving skill and workflow content and create long-term synchronization overhead.
Additional context
#438 / #439 already established a small precedent for improving compatibility with another Agent Skills runtime (GitHub Copilot CLI) while preserving Claude Code behavior.
Before going further, I'd mainly like to know:
Contribution