Antigravity-Master-Prompt-Source

A reusable AI Agent Operating System for Google Gemini Antigravity IDE coding and technical assistant workflows.

This repository provides modular prompt infrastructure that helps an AI agent consistently select the right role, apply safety guardrails, follow implementation playbooks, and produce disciplined outputs.

Repository Structure

Antigravity-Master-Prompt-Source/
├── AGENTS.md
├── routers/
│   └── prompt-router.md
├── roles/
│   ├── senior-software-engineer.md
│   ├── senior-audio-sound-engineer.md
│   ├── senior-ui-designer.md
│   ├── senior-project-manager.md
│   ├── robotics-engineer.md
│   └── desktop-engineer-helpdesk.md
├── guardrails/
│   ├── ui-preservation.md
│   ├── security-first-development.md
│   └── repo-discipline.md
├── playbooks/
│   ├── project-start-checklist.md
│   ├── debugging-playbook.md
│   ├── deployment-checklist.md
│   └── repo-janitor-workstreams.md
└── bundles/
    ├── software-dev-stack.md
    ├── robotics-stack.md
    └── helpdesk-stack.md

Core Concepts

• AGENTS.md: Master behavior controller and execution workflow.
• Router: Maps tasks to role modules.
• Roles: Professional prompt definitions with responsibilities and expected outputs.
• Guardrails: Non-negotiable policies (UI preservation, security, repo discipline).
• Playbooks: Repeatable operational checklists for start, debug, deploy, and repo hygiene.
• Bundles: Curated combinations of roles + guardrails by scenario.

How to Use

1. Start with AGENTS.md.
2. Route task intent using routers/prompt-router.md.
3. Load role module(s) from roles/.
4. Apply relevant guardrails from guardrails/.
5. Execute using a playbook from playbooks/.
6. Optionally load a prebuilt stack from bundles/.

The current prompt system is optimized for:

• minimal-diff implementation
• strong UI and behavior preservation
• security-first decisions
• branch and workstream discipline
• beginner-safe explanations when needed
• low-friction reuse across multiple project types

Status

Initialized with core operating modules for:

• software/web delivery
• audio and interactive sound-system work
• UI-sensitive tasks
• robotics/electronics work
• desktop/helpdesk troubleshooting