Working with AI (Antigravity)
At PlugZero Analytics, we use the Antigravity AI Agent as a core part of our team. This isn’t just a chatbot; it is a tool that has full access to our code and tools.
How the Agent helps
The agent is trained to follow our specific rules. It helps in three main ways:
1. Building Features
When you ask the agent to build a new view or endpoint, it will:
- Read your existing code to find the right patterns.
- Write the logic.
- Run the server to make sure there are no errors.
2. Writing Documentation
This entire documentation portal was built by the agent.
- Every time we change a major part of the math engine or the UI, the agent is responsible for updating the Registry pages.
- It uses the “Platform Documenter” skill to scan the code and write these descriptions.
3. Fixing Bugs
The agent can read terminal errors (like Next.js build errors) and fix the config files automatically.
How to use the Agent correctly
To get the best results from the AI agent, follow these steps:
Be specific
Instead of saying “Fix the login,” say “Update the LoginView in the backend to handle 2FA tokens.”
Let it research
The agent has a “Technical Researcher” skill. If you are starting a new task, ask it to “Research the existing ingestion logic” before it writes any code.
Verify its work
The agent is good, but it can make mistakes. Always check the file diffs (changes) it makes before you commit the code to Git.
Security Rule: Never give the agent your personal passwords or private API keys. Use environment variables (like .env files) instead.