Authoring Google Antigravity Skills

This repository is the working project for a codelab on authoring Google Antigravity Skills.

This codelab uses a practical throughline: build a tool to clean event registration data stored in CSV files. The use case is easy to understand, maps well to real operational work, and makes the value of reusable skills immediately obvious.

Codelab Overview

Participants work with a messy event registration file and use Google Antigravity to progressively improve how that work gets done.

The session is organized around two focused parts:

1. Use a few sample/common skills on a concrete problem Participants explore the event registration dataset, review inconsistent values, and use Antigravity capabilities and existing workflows on a task that feels realistic rather than abstract.
2. Prompt Antigravity to create a custom skill for that problem Participants then author a custom skill that supports the CSV-cleaning workflow directly, turning an ad hoc prompt into a reusable capability.

This structure fits the product behavior well: Skills are loaded on demand through intent matching from their descriptions, so it is natural to first show how existing skills are invoked and then how to create a new one tailored to the exact task.

What Participants Learn

By the end of the codelab, participants should understand:

• why Skills matter in Google Antigravity
• how on-demand skill loading works
• how a SKILL.md file is structured
• when to scope a skill globally versus at the workspace level
• how to use prompt-based skill authoring to create a task-specific workflow

They also get hands-on exposure to the broader Antigravity working model, including:

• Agent Manager
• Editor workflows
• Planning mode versus Fast mode
• terminal review and iteration
• Rules and Workflows

Practical Scenario

The repository includes a sample CSV dataset: registration_raw.csv.

It represents event registration data with intentionally messy values.

This makes the codelab concrete: participants are not learning Skills in the abstract, they are using Skills to solve a recognizable data-cleaning problem.

Outcome

At the end of the session, participants should leave with two things:

1. practical experience using Antigravity on a concrete task
2. a clear mental model for turning that task into a reusable custom Skill

---

Developed by Minh N. Ta.