Your First Gate Moment

This page walks through the experience of Anvil catching an issue in real-time.

The Scenario

You're using an AI coding assistant to add a new endpoint. The AI generates working code, but it violates your architecture.

Step 1: Start Watch Mode

In a terminal:

anvil watch

You see:

🔨 Anvil Watch

Watching for changes...
Press Ctrl+C to stop.

Step 2: AI Generates Code

You ask your AI assistant: "Add a DELETE endpoint for users"

The AI generates src/api/handlers/delete-user.ts:

import { Request, Response } from 'express';
import { db } from '../../repositories/db'; // Direct DB access!

export async function deleteUser(req: Request, res: Response) {
  const { id } = req.params;

  try {
    await db.query('DELETE FROM users WHERE id = $1', [id]);
    res.status(204).send();
  } catch (e) {
    // AI left an empty catch block
    res.status(500).send();
  }
}

Step 3: Save the File

The moment you save, Anvil responds:

🔨 Change detected: src/api/handlers/delete-user.ts

Checking architecture...
  ✗ ARCH-001: Boundary violation
    src/api/handlers/delete-user.ts:2
    imports from ../../repositories/db
    Rule: api-layer denies imports from src/repositories/**

    API handlers should use services, not repositories directly.

Checking anti-patterns...
  ⚠️  AP-006: Empty catch block
      src/api/handlers/delete-user.ts:10:5

      Empty catch blocks hide errors. Log the error or re-throw.

1 error, 1 warning found.
Gate status: FAIL

Step 4: Fix Before Commit

You now know immediately that this code has issues—before you commit, before you push, before a reviewer has to point it out.

Fix the architecture violation:

import { Request, Response } from 'express';
import { UserService } from '../../services/user.service'; // Correct!

export async function deleteUser(req: Request, res: Response) {
  const { id } = req.params;

  try {
    await UserService.delete(id);
    res.status(204).send();
  } catch (error) {
    console.error('Failed to delete user:', error); // Proper handling
    res.status(500).json({ error: 'Failed to delete user' });
  }
}

Save again:

🔨 Change detected: src/api/handlers/delete-user.ts

Checking architecture... ✓
Checking anti-patterns... ✓

All gates passed.

The Value

In traditional workflows:

1. AI generates code
2. You commit and push
3. CI runs (5 minutes)
4. Reviewer spots the issue (hours later)
5. You context-switch back to fix

With Anvil:

1. AI generates code
2. You save
3. Anvil catches it (milliseconds)
4. You fix while context is fresh

Time saved: Hours of review cycles, context-switching, and accumulated technical debt.

What Gates Provide

Traditional	With Anvil
Issues found in review	Issues found at save
Reviewer cognitive load	Automated enforcement
Inconsistent enforcement	Deterministic rules
"It passed tests" excuses	Architecture is tested too
Invisible AI drift	Visible, immediate feedback

---

Next: Understand the concepts behind gates →