AI Toolkit

Reusable skills, agents, prompts, templates, and workflow helpers for AI-assisted projects.

This repo is intentionally broad. It started with OpenSCAD and MakerWorld skills, but it is set up to hold future project helpers that are not tied to a single codebase.

Layout

• skills/: reusable agent skills with SKILL.md files and nearby references.
• agents/: agent role definitions, operating notes, or reusable persona/workflow specs.
• prompts/: prompts that are useful across projects.
• templates/: reusable file templates, issue/PR text, checklists, or starter content.
• scripts/: helper scripts used by skills or workflows.
• docs/: notes about how to use and maintain the toolkit.

Current Skills

• skills/openscad-authoring: general OpenSCAD structure, modular modeling, parameter organization, and maintainable source layout.
• skills/makerworld-openscad: MakerWorld-specific OpenSCAD export rules.
• skills/makerworld-listing: MakerWorld listing copy, README text, customization guides, tags, and release notes.
• skills/makerworld-images: MakerWorld listing images, covers, gallery sequencing, comparison images, and GIFs.

Install

This repo is compatible with the skills CLI because skills live under skills/ and each skill has a SKILL.md file.

List available skills:

npx skills add ejwill/ai-toolkit --list

Install all skills globally for Codex:

npx skills add ejwill/ai-toolkit --skill '*' --global --agent codex --yes

Install one skill globally for Codex:

npx skills add ejwill/ai-toolkit --skill openscad-authoring --global --agent codex --yes

By default the CLI uses symlinks when possible, which keeps installed skills tied back to one canonical copy.

Conventions

Each skill should live in its own directory:

skills/
  example-skill/
    SKILL.md
    reference.md
    scripts/
    examples/

Keep SKILL.md short and operational. Put longer explanations, examples, and checklists in adjacent reference files so agents can load only what they need.

Use lowercase kebab-case for skill directories and skill names. Treat product names as one token when that is how the project writes them, such as openscad rather than open-scad.