Antigravity skill creator
Skill Babalsaab/Project_Starter/.claude/skills/antigravity-skill-creator
Use when creating skills compatible with AntiGravity agent environment, or when building cross-platform skills that work in both Claude Code and AntiGravity/.agent/ setups.From its SKILL.md
npx -y skills add Babalsaab/Project_Starter --skill antigravity-skill-creatorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.3 KB, as published. Nobody here has run it
AntiGravity Skill Creator
Creates skills compatible with the AntiGravity agent environment (.agent/skills/) following the Agent Skills open standard. Skills created here also work with Gemini CLI, Cursor, Windsurf, and other agents supporting the standard.
When to Use
- Building a skill for AntiGravity or
.agent/skills/format - Creating cross-platform skills (Claude Code + AntiGravity + others)
- Converting existing skills to the Agent Skills open standard
Folder Structure
Every skill must follow this hierarchy:
<skill-name>/
SKILL.md # Required: Main logic and instructions (<500 lines)
scripts/ # Optional: Helper scripts (validation, API calls)
examples/ # Optional: Reference implementations
resources/ # Optional: Templates, checklists, assets
YAML Frontmatter Standards
---
name: gerund-form-name # e.g., "testing-code", "managing-databases"
description: Third person. # Include specific trigger keywords. Max 1024 chars.
allowed-tools: Read, Write # Comma-separated tool restrictions
---
Rules:
- name: Gerund form, max 64 chars, lowercase + numbers + hyphens only
- description: Written in third person with specific triggers/keywords
- No "claude" or "anthropic" in the name
Writing Principles (The "Claude Way")
- Conciseness — Assume the agent is smart. Don't explain basics. Focus on unique logic.
- Progressive Disclosure — Keep SKILL.md under 500 lines. Link to secondary files only one level deep.
- Forward Slashes — Always use
/for paths, never\. - Degrees of Freedom:
- Bullet Points — High-freedom tasks (heuristics, guidelines)
- Code Blocks — Medium-freedom (templates, schemas)
- Specific Bash Commands — Low-freedom (fragile operations that must be exact)
Workflow & Feedback Loops
For complex tasks, include:
- Checklists — Markdown checklist the agent copies and updates to track state
- Validation Loops — Plan-Validate-Execute pattern (check config BEFORE applying changes)
- Error Handling — Scripts are "black boxes" — tell agent to run
--helpif unsure
Cross-Platform Skill Creation
To create a skill that works in BOTH Claude Code and AntiGravity:
Step 1: Build the AntiGravity version
.agent/skills/<name>/
SKILL.md
scripts/
resources/
Step 2: Symlink or copy to Claude Code
.claude/skills/<name>/
SKILL.md # Same file, Claude Code frontmatter is a superset
Step 3: Add Orchestrator integration (optional)
skills/<name>/
config.yaml # Orchestrator-specific: compatible_agents, domain, priority
SKILL.md # Same content
Frontmatter Compatibility
| Field | Claude Code | AntiGravity | Both |
|---|---|---|---|
name | Yes | Yes | Yes |
description | Yes | Yes | Yes |
allowed-tools | Yes | Yes | Yes |
context: fork | Yes | No | Claude only |
model | Yes | No | Claude only |
hooks | Yes | No | Claude only |
Skill Template
When creating a new skill, use this template:
---
name: [gerund-name]
description: [3rd-person description with trigger keywords]
allowed-tools: [tools needed]
---
# [Skill Title]
## When to Use This Skill
- [Trigger scenario 1]
- [Trigger scenario 2]
## Workflow
1. [Step with clear instruction]
2. [Step with validation]
3. [Step with output]
## Instructions
[Specific logic, code snippets, or rules]
## Checklist
- [ ] [Verification item 1]
- [ ] [Verification item 2]
## Resources
- [Link to scripts/ or resources/ if applicable]
Output Convention
When generating a skill, output:
- The full directory structure
- Complete SKILL.md content
- Any supporting files (scripts, resources, examples)
- Installation instructions for the target platform
Notes
- Test both
/skill-nameinvocation and natural language triggering - Verify skills don't exceed the character budget (2% of context window)
- For skills with side effects (API calls, file generation), set
disable-model-invocation: true