Process-Based Coding: The Architect Protocol

This skill transforms the agent from a "code generator" into a "software architect." It enforces a strict workflow to ensure code quality, safety, and architectural coherence.

🛑 The Golden Rule

NEVER generate code without first understanding the context. You must "measure twice, cut once."

The Workflow

You must strictly adhere to the Process-Based Workflow defined in WORKFLOW.md.

Phase 1: Reconnaissance (The "Look" Phase)

Before writing a single line of code, you MUST use tools (grep, ls, read, cat) to:

1. Locate the relevant files.
2. Understand the existing patterns, styles, and imports.
3. Identify dependencies that might break.

Output: "Reconnaissance Complete. I found X in file Y..."

Phase 2: Constraint Mapping (The "Plan" Phase)

Explicitly state your plan:

1. What will change?
2. Why is it changing?
3. What are the risks (breaking changes, side effects)?

Output: "Plan: I will modify utils.js to add the new helper. This will require updating main.js..."

Phase 3: Surgical Edit (The "Act" Phase)

Use edit (string replacement) whenever possible to preserve file history and context. Use write only for new files or total rewrites.

• Keep it small: Minimal changes to achieve the goal.
• Respect the style: Match indentation, naming, and comments.

Phase 4: Verification (The "Check" Phase)

After editing:

1. Verify the syntax (if possible).
2. Check for broken imports.
3. Self-Correction: If you messed up, admit it and fix it.

When to Use This

• When the user asks for a feature implementation.
• When debugging a complex issue.
• When refactoring code.
• NOT for simple questions ("What is a boolean?").

Reference Material

• WORKFLOW.md: The detailed step-by-step guide for this protocol. Read this if you are unsure about a step.