Skill Creator Skill

This skill governs the creation of other Antigravity Skills. It ensures consistency, best practices, and high-quality instruction sets.

Core Principles

1. •Atomic Purpose: Each skill should do one thing exceptionally well.
2. •Explicit Instructions: Do not assume the agent knows specific project context.
3. •Progressive Disclosure: Push heavy logic into scripts or templates; keep SKILL.md for high-level direction.

Directory Structure

Always follow this structure when creating a new skill:

<skill-name>/
├── SKILL.md        (Required: The brain)
├── scripts/        (Optional: The hands - Python/Bash/Node)
├── references/     (Optional: The knowledge - Docs/Templates)
└── examples/       (Optional: The training - Inputs/Outputs)

SKILL.md Template

Every SKILL.md must start with this YAML frontmatter:

---
name: <skill-name-kebab-case>
description: <Trigger phrase>. Use this when <specific condition>.
---

Sections to Include

• •Goal: What does this skill achieve?
• •Instructions: Step-by-step logic.
• •Constraints: What NOT to do (e.g., "Do not delete files").
• •Example Usage: A brief conversation snippet showing the user prompt and expected agent action.

Best Practices

• •Scripts: Prefer Python for logic. Always provide a clear usage string in scripts.
• •Dependencies: If a script needs libraries, check if they are standard. If not, instruct the agent to install them or check availability.
• •Validation: When creating a skill, always instruct the agent to test it immediately after creation if possible.