Skip to content
Available on:
Antigravity 2.0Antigravity CLIAntigravity IDE

Skills are an open standard for extending agent capabilities. A skill is a folder containing a SKILL.md file with instructions that the agent can follow when working on specific tasks.

Skills are reusable packages of knowledge that extend what the agent can do. Each skill contains:

  • Instructions: explicit protocols for how to approach a specific task.
  • Best practices: conventions, style guidelines, and checklists to follow.
  • Scripts and resources: optional helper scripts and data schemas the agent can execute.

When you start a conversation, the agent sees a list of available skills with their names and descriptions. If a skill looks relevant to your task, the agent reads the full instructions and follows them.

Skills are organized as directory bundles containing a required SKILL.md file:

my-skill/
├── SKILL.md       # Required skill instructions and metadata
├── scripts/       # Optional executable scripts
├── examples/      # Optional reference implementations
└── resources/     # Optional templates, schemas, or data files

Every SKILL.md file begins with YAML frontmatter defining its name and triggering criteria:

---
name: code-review
description: Reviews code changes for bugs, style issues, and best practices. Use when reviewing pull requests or checking code quality.
---

# Code Review Skill

When reviewing code, follow these steps:

## Review checklist

1. **Correctness**: verify that the code satisfies specifications.
2. **Edge cases**: ensure error conditions and boundaries are handled.
3. **Style**: follow project naming and architectural patterns.
4. **Performance**: identify potential bottlenecks or inefficiencies.

The YAML frontmatter supports the following fields:

FieldRequiredDescription
nameNoA unique identifier for the skill (lowercase, hyphens for spaces). Defaults to the folder name if not provided.
descriptionYesA clear description of what the skill does and when to use it. This is what the agent sees when deciding whether to apply the skill.

Skills follow a progressive disclosure pattern:

  1. Discovery: when a conversation starts, the agent sees a list of available skills with their names and descriptions.
  2. Activation: if a skill looks relevant to your task, the agent reads the full SKILL.md content.
  3. Execution: the agent follows the skill’s instructions while working on your task.

You don’t need to explicitly tell the agent to use a skill—it decides based on context. However, you can mention a skill by name if you want to ensure it’s used.

Each skill should do one thing well. Instead of a “do everything” skill, create separate skills for distinct tasks.

The description is how the agent decides whether to use your skill. Make it specific about what the skill does and when it’s useful.

If your skill includes scripts, encourage the agent to run them with --help first rather than reading the entire source code. This keeps the agent’s context focused on the task.

For complex skills, add a section that helps the agent choose the right approach based on the situation.

Explore how to create and manage skills on your preferred surface:

Antigravity 2.0 loads skills from two primary locations:

LocationScope
<workspace-root>/.agents/skills/<skill-folder>/Workspace-specific
~/.gemini/config/skills/<skill-folder>/Global (all workspaces)
  • Workspace skills: scoped to a single project and committed to version control to share across your engineering team.
  • Global skills: available across all projects on your workstation.
  • Autonomous invocation: the agent automatically reads and follows relevant skills based on your prompt.
  • Manual slash command: type /<skill-name> in the prompt panel to explicitly invoke a skill.