Ready-to-use templates for wiring these skills into AI platforms.
| File | Use with | What it gives you |
|---|---|---|
| Claude Code plugin | Claude Code | /plugin marketplace add shoka-jp/automotive-ai-skills → all 20 skills install natively (SKILL.md per skill) and auto-trigger |
| codex.md | OpenAI Codex | AGENTS.md auto-load in-repo, global setup, or one slash command per skill |
| claude-project-template.md | Claude Projects, Custom GPTs, or any system-prompt slot | A single-skill dedicated assistant, step by step |
| garage-assistant-system-prompt.md | Any assistant with a system prompt | One combined "garage assistant" that routes between all skills |
skill.yaml files (one per skill folder) |
Agent frameworks, RAG pipelines, skill registries | Machine-readable skill metadata: name, summary, tags, languages, related skills |
- Just chatting? You don't need any of this — copy an example prompt from a skill's README and go.
- Repeated use of one skill (e.g., you're mid-restoration and consult the Classic Car Restorer weekly): use the Claude Project template with that one skill.
- General-purpose garage helper for a household or shop: use the combined system prompt.
- Building software (an agent, a RAG bot, a fleet tool): load the
skill.yamlmetadata for routing, and the README content as the knowledge payload. All files are MIT-licensed.
skill.yamlschema is documented in the Skill Authoring Guide.- Each skill's
Limitationssection is the safety-relevant part — if you compress skill content to fit a context budget, keep Limitations intact. - Japanese content (
README.ja.md,docs/ja/) lets you build a Japanese-first assistant with the same structure.