skill-antigravity

これはレビュー時に保存された抜粋です。完全で最新の内容は外部ソースを確認してください。

---
name: skill-antigravity
description: Provides core knowledge and instructions for creating, documenting, and managing Antigravity Skills. Use this skill when you need to extend your capabilities by creating new modular skills.
version: 1.0.0
author: Muze AI
license: MIT
---

# Antigravity Skills Master

This skill empowers Antigravity agents to extend their own capabilities through the creation and management of modular skills.

## Core Principles
1. **Modularity**: Each skill should focus on a single, well-defined task or domain.
2. **Context Efficiency**: Keep instructions concise to minimize token usage while maintaining clarity.
3. **Discoverability**: Use precise, keyword-rich descriptions in the frontmatter to help the agent identify relevance.
4. **Reliability**: Provide checklists and error-handling logic for critical tasks.
5. **Singular Naming**: Always use `.agent` (singular), never `.agents` (plural), for the root configuration directory to ensure automatic tool detection.

## Anatomy of a Skill
A standard Antigravity skill follows this structure:
- `SKILL.md`: (Required) The core instruction set with YAML metadata.
- `metadata.json`: (Recommended) Structured metadata for the `skills` CLI.
- `scripts/`: (Optional) Automation scripts or helper tools.
- `examples/`: (Optional) Reference implementations.
- `resources/`: (Optional) Static assets, templates, or rulesets.

## Guidelines for Creating Skills
1. **Identify the Need**: Create a skill when a task is repetitive, complex, or requires specialized domain knowledge.
2. **Setup Folder**: Create a directory in `.agent/skills/` (local) or your skills repository.
3. **Define Metadata**: Ensure the `name` matches the folder name and the `description` is actionable.
4. **Draft Instructions**:
   - Use clear sections (Goal, Requirements, Steps, Verification).
   - Include GitHub-style alerts for critical warnings or tips.
   - Add Mermaid diagrams for complex workflows.

## Managing the Skills Ecosystem
- **Local Workspace**: Store project-specific skills in `.agent/skills/`.
- **Global Library**: Use `npx skills add` to import skills from shared repositories.
- **Versioning**: Update the `version` in frontmatter/metadata when making significant changes.

## Verification Checklist
- [ ] Does the folder name match the `name` in `SKILL.md`?
- [ ] Is the `description` clear enough for discovery?
- [ ] Are all external dependencies (scripts) referenced correctly?
- [ ] Has the skill been tested in a real-world scenario?