agent-guardrails-template

# Agent Guardrails Template

> AI-first safety framework for agents building software at high velocity. Guardrails don't slow you down — they're your license to move fast.

[![Version](https://img.shields.io/badge/version-v2.8.0-blue.svg)](./CHANGELOG.md)
[![Go Implementation](https://img.shields.io/badge/Implementation-Go-blue.svg?style=flat&logo=go)](https://golang.org)
[![WCAG 3.0+](https://img.shields.io/badge/Accessibility-WCAG_3.0+_Silver-green.svg)](docs/accessibility/ACCESSIBILITY_GUIDE.md)
[![Spatial Computing](https://img.shields.io/badge/Spatial-XR/VR/AR-blue.svg)](docs/spatial/SPATIAL_COMPUTING_UI.md)

---

## What Is This?

**The Agent Guardrails Template** is a framework that enables AI agents to develop software at full velocity with built-in safety. Whether you're vibe coding a game prototype or building production infrastructure, guardrails let agents spend tokens on building instead of second-guessing safety.

It works with any AI system — Gemini, GPT, Gemini, LLaMA, and 30+ other model families.

### The Paradox: Constraints Enable Speed

Without guardrails, agents waste tokens on safety verification: *"Is this file safe to edit? Will this break something? Should I ask first?"* This constant self-checking burns context and slows output.

With guardrails, agents know the boundaries upfront. They spend tokens on building, not on doubt. The result: faster iteration, fewer rollbacks, and code that ships with confidence.

Think of guardrails like lane markers on a highway — they don't slow you down. They're the reason you can drive at full speed.

### The Four Laws of Agent Safety

1. **Read before editing** — Never modify code without reading it first
2. **Stay in scope** — Only touch files explicitly authorized
3. **Verify before committing** — Test and check all changes
4. **Halt when uncertain** — Ask for clarification instead of guessing

---

## Quick Start

```bash
# Clone the template
git clone https://github.com/TheArchitectit/agent-guardrails