Creates new Antigravity skills with streamlined workflows and automation. Use when users want to create a new skill, improve an existing skill, or need guidance on skill structure. Triggers include "create a skill", "make a new skill", "build a skill for", "skill for [task]", "help me create a skill", or any request to extend Antigravity's capabilities with specialized knowledge or workflows.
Installation
Details
Usage
After installing, this skill will be available to your AI coding assistant.
Verify installation:
npx agent-skills-cli listSkill Instructions
name: skill-generator description: Creates new Antigravity skills with streamlined workflows and automation. Use when users want to create a new skill, improve an existing skill, or need guidance on skill structure. Triggers include "create a skill", "make a new skill", "build a skill for", "skill for [task]", "help me create a skill", or any request to extend Antigravity's capabilities with specialized knowledge or workflows.
Skill Generator
Create effective Antigravity skills quickly with clear workflows, automation scripts, and ready-to-use templates.
Quick Start (5 Steps)
Get from idea to working skill in 10 minutes:
- Initialize: Run
python scripts/init_skill.py <skill-name> --path .agent/skills - Edit frontmatter: Update
nameanddescriptionin SKILL.md (critical for triggering) - Write instructions: Add your skill's core guidance (use imperative form)
- Clean up: Delete unused example files from
scripts/,references/,assets/ - Package: Run
python scripts/package_skill.py .agent/skills/<skill-name>
Done! You now have a .skill file ready to share.
Skill Creation Workflow
Step 1: Understand the Need
Before creating a skill, clarify:
- What task does this skill enable?
- When should it trigger? (specific user requests, file types, scenarios)
- What makes it reusable? (repeated workflows, specialized knowledge, automation)
Example questions to ask:
- "What functionality should this skill support?"
- "Can you give examples of how it would be used?"
- "What would a user say to trigger this skill?"
Step 2: Plan Reusable Contents
Analyze what resources would help when executing this task repeatedly:
Scripts (scripts/) - Use when:
- Same code is rewritten repeatedly
- Deterministic reliability needed
- Complex operations (PDF manipulation, API calls, data processing)
References (references/) - Use when:
- Detailed documentation needed (API docs, schemas, policies)
- Information too lengthy for SKILL.md
- Content only needed for specific use cases
Assets (assets/) - Use when:
- Templates or boilerplate needed (PowerPoint, HTML, React projects)
- Images, fonts, or media files required
- Files used in output (not loaded into context)
Instruction-only - Use when:
- Workflow guidance is sufficient
- No repeated code or templates
- Simple procedural knowledge
Step 3: Initialize the Skill
Run the initialization script:
python scripts/init_skill.py <skill-name> --path .agent/skills
Naming requirements:
- Hyphen-case (e.g.,
pdf-editor,nz-job-search) - Lowercase letters, digits, hyphens only
- Max 40 characters
- Must match directory name
This creates:
.agent/skills/<skill-name>/
βββ SKILL.md (with TODO placeholders)
βββ scripts/example.py
βββ references/api_reference.md
βββ assets/example_asset.txt
Step 4: Write SKILL.md
Frontmatter (Critical)
---
name: skill-name
description: [Complete description of what the skill does AND when to use it]
---
Description best practices:
- Include WHAT the skill does
- Include WHEN to trigger (specific scenarios, file types, keywords)
- List example trigger phrases
- Be comprehensive - this determines when the skill loads
Example:
description: Creates and edits PDF documents with support for forms, annotations, and text extraction. Use when working with PDF files for: (1) Filling forms, (2) Extracting text, (3) Merging/splitting, (4) Adding annotations. Triggers: "edit this PDF", "fill PDF form", "extract text from PDF", "merge PDFs".
Body Structure
Choose a structure that fits your skill:
Workflow-Based (sequential processes):
## Overview
## Workflow Decision Tree
## Step 1: [Action]
## Step 2: [Action]
Task-Based (tool collections):
## Overview
## Quick Start
## Task 1: [Capability]
## Task 2: [Capability]
Reference-Based (standards/guidelines):
## Overview
## Guidelines
## Specifications
## Usage Examples
Writing guidelines:
- Use imperative form ("Create", "Run", "Check" not "Creating", "You should")
- Keep SKILL.md under 500 lines (split into references if longer)
- Include concrete examples
- Reference bundled resources clearly
- Delete the template guidance sections
Step 5: Add Bundled Resources
Scripts
Test all scripts by running them:
python scripts/your_script.py
Ensure they work without errors. Scripts should be self-contained and documented.
References
For files >100 lines, add a table of contents at the top. Structure references for easy scanning.
Progressive disclosure: Link to references from SKILL.md only when needed:
For advanced features, see [advanced.md](references/advanced.md)
Assets
Store templates, images, fonts, or boilerplate that will be used in output. These are NOT loaded into context.
Delete unused directories - Not every skill needs all three types.
Step 6: Validate & Package
Before packaging, validate:
python scripts/quick_validate.py .agent/skills/<skill-name>
Fix any errors, then package:
python scripts/package_skill.py .agent/skills/<skill-name>
This creates <skill-name>.skill - a distributable zip file.
Validation checklist:
- β
YAML frontmatter has
nameanddescription - β Description includes "when to use" triggers
- β Instructions use imperative form
- β Unused example files deleted
- β Scripts tested and working
- β SKILL.md under 500 lines
- β References linked from SKILL.md
Skill Design Principles
1. Concise is Key
Context window is shared. Only include what Claude doesn't already know.
Challenge each section: "Does Claude really need this?" and "Does this justify its token cost?"
Prefer examples over explanations.
2. Progressive Disclosure
Skills load in three levels:
- Metadata (name + description) - Always in context
- SKILL.md body - When skill triggers
- Bundled resources - As needed
Keep SKILL.md lean. Move detailed content to references.
3. Set Appropriate Freedom
Match specificity to task fragility:
- High freedom (text instructions): Multiple valid approaches
- Medium freedom (pseudocode): Preferred pattern with variation
- Low freedom (specific scripts): Fragile operations requiring exact steps
Templates & Examples
Ready-to-Use Templates
See references/skill-templates.md for copy-paste templates:
- Instruction-only skill
- Script-based skill
- Reference-heavy skill
- Tool integration skill
Real Skill Examples
See references/skill-examples.md for complete, annotated examples showing different patterns.
Design Patterns
See existing reference docs for proven patterns:
- references/workflows.md - Sequential workflows and conditional logic
- references/output-patterns.md - Template and example patterns
Common Patterns
Multi-Framework Skills
When supporting multiple frameworks/variants, organize by variant:
skill-name/
βββ SKILL.md (overview + selection guide)
βββ references/
βββ framework-a.md
βββ framework-b.md
βββ framework-c.md
Claude loads only the relevant reference when the user chooses a framework.
Domain-Specific Organization
For skills with multiple domains:
bigquery-skill/
βββ SKILL.md (navigation)
βββ references/
βββ finance.md
βββ sales.md
βββ product.md
Claude loads only the relevant domain reference.
Conditional Details
Show basic content, link to advanced:
## Basic Usage
[Simple instructions]
**For advanced features**: See [advanced.md](references/advanced.md)
Troubleshooting
Skill not triggering?
- Check
descriptionincludes trigger phrases - Add more "when to use" scenarios
- Include specific keywords users might say
SKILL.md too long?
- Move detailed content to
references/ - Keep only essential workflow in SKILL.md
- Use progressive disclosure pattern
Scripts not working?
- Test scripts directly before packaging
- Ensure all dependencies documented
- Add error handling and clear output
Validation failing?
- Check YAML frontmatter format
- Ensure
namematches directory name - Verify
descriptionis comprehensive
What NOT to Include
Do not create extraneous documentation:
- README.md
- INSTALLATION_GUIDE.md
- CHANGELOG.md
- QUICK_REFERENCE.md
Skills are for AI agents, not humans. Include only what's needed for the task.
Iteration
After using the skill:
- Notice struggles or inefficiencies
- Identify needed improvements
- Update SKILL.md or resources
- Re-package and test
Skills improve through real usage.
More by EricZhou0815
View allinterview-prep-coach: Helps candidates prepare strategically for job interviews by simulating hiring manager thinking, identifying risks, generating likely questions, and coaching strong structured answers. Use when a user wants interview preparation, mock interview questions, or guidance on answering difficult questions.
Crafts strategic, high-impact cover letters that address Hiring Manager risks and bridge qualification gaps. Use when users need a cover letter for a specific job, want to explain a career gap/pivot, or need to position themselves for a role they are slightly underqualified for. Triggers include "write a cover letter", "draft an application letter", "explain my gap in a letter", "apply for this job".
Finds current New Zealand software engineering leadership jobs matching specific criteria. Use when searching for NZ tech jobs, leadership roles in Christchurch or remote NZ positions, or when the user asks to "find NZ jobs", "search for tech lead roles in NZ", "find Christchurch software jobs", "search for NZ engineering manager positions", or similar job search requests focused on New Zealand.
name: application-coach description: Evaluates a candidate's CV and
