[Feature] Integrate Claude skill-creator to automate skill creation

[Y0uYuGe]

Summary

Integrate Claude's skill-creator workflow as an optional auxiliary capability to automate skill creation, so users can generate a reusable skill package from natural language requirements, existing scripts, or project workflows instead of authoring every skill manually.

Motivation or Use Case

Today, creating a skill usually requires manual work such as:

• writing SKILL.md
• defining usage boundaries and instructions
• organizing scripts and supporting files
• deciding input / output conventions
• preparing examples and metadata

This is manageable for a few skills, but becomes costly when:

• a team wants to convert repeated workflows into reusable skills
• users want to quickly turn a prompt / script / local procedure into a skill
• a repository wants to accumulate skills as reusable assets rather than ad hoc instructions
• contributors are unfamiliar with the expected skill structure

A skill-creator integration would reduce friction and make skill authoring more scalable.

Who benefits:

• users who want to quickly create a new skill from a task description
• maintainers who want more consistent skill structure
• teams converting internal workflows into shared reusable skills
• ecosystems that want to grow a larger and more standardized skill library

Proposed Change

Add support for automated skill generation via Claude skill-creator, either through direct integration or through a compatible adapter layer.

1. Creation sources

Allow users to generate a skill from one or more of the following inputs:

• a natural language task description
• an existing prompt or workflow description
• a local script or command sequence
• a project directory or example files
• an existing skill that needs refinement or normalization

2. Generated outputs

The automated creation flow should be able to produce an initial skill package that may include:

• SKILL.md
• usage instructions
• boundaries / constraints
• input / output expectations
• supporting scripts or wrappers
• examples or usage hints
• optional tags / metadata for indexing and retrieval

3. Suggested workflow

Possible flow:

1. user provides a task goal or source material
2. system invokes skill-creator
3. system generates a draft skill package
4. user reviews / edits the generated result
5. skill is saved locally or prepared for upload

This should produce a draft, not silently finalize everything without review.

4. CLI integration

Possible command ideas:

• magicskills create-skill
• magicskills create-skill --from-prompt
• magicskills create-skill --from-script
• magicskills create-skill --from-dir
• magicskills refine-skill <path>

The exact command names can vary, but the main goal is to support an automated authoring path rather than only manual creation.

5. Review and refinement

Generated skills should be easy to inspect and refine before use.
For example:

• preview generated SKILL.md
• review referenced commands
• adjust metadata
• edit examples
• optionally run scan / validation before saving or publishing

6. Validation compatibility

This feature should work well with future or existing:

• skill validation
• retrieval indexing
• security scanning
• upload / publish workflows

The main idea is that auto-generated skills should still fit into the same lifecycle as manually authored skills.

Alternatives Considered

Alternative 1: Manual skill authoring only

This gives maximum control, but has high friction and slows down skill accumulation.

Alternative 2: Provide templates only

Templates help, but users still need to manually translate task knowledge into the expected skill structure.

Alternative 3: Build a project-native generator from scratch

This offers full control, but increases implementation cost and duplicates work if a useful external skill-creator already exists.

Preferred direction

A Claude skill-creator integration or compatible adapter provides the fastest path to automated skill authoring while keeping room for project-specific customization later.

Documentation or Example Impact

The following documentation would likely need updates:

• skill authoring guide
• CLI reference
• examples of creating a skill from a prompt
• examples of creating a skill from a script or workflow
• review checklist for generated skills
• best practices for refining auto-generated skills

It would also help to add:

• before / after examples
• common input patterns that generate better skills
• limitations of automatically generated skill packages

Additional Context

This feature is important for treating skills as reusable assets rather than manually assembled folders.

A good automated creator can help:

• standardize skill structure
• reduce authoring cost
• improve ecosystem growth
• turn repeated workflows into reusable capabilities faster

At the same time, generated skills should not be assumed to be correct or safe by default.
They should remain reviewable, editable, and ideally compatible with:

• validation
• security scanning
• retrieval indexing
• upload / publishing controls

A future extension could include:

• interactive skill refinement
• generate + scan workflow
• generate + test workflow
• converting existing project workflows into skills in batch
• converting skills between different agent ecosystems