context-not-control

# Context, not Control

A skill that transforms how you work with AI - from micromanaging every step to providing context and letting AI make decisions. Inspired by the "Context, not Control" philosophy from the OpenClaw community.

## Core Philosophy

**Traditional approach**: You tell AI exactly what to do, step by step.  
**This approach**: You tell AI what you want to achieve, AI figures out how.

The key insight: AI works best when you give it rich context about your goals, constraints, and preferences - then trust it to execute within appropriate boundaries.

## When to Use This Skill

- Starting a new project with vague requirements
- Want to reduce back-and-forth and rework
- Need AI to take more initiative and make decisions
- Want to establish clear permission boundaries
- Transitioning from "micromanaging AI" to "trusting AI"

## Quick Start

### 1. Initialize Your Context

Run the initialization script to set up your project context and permission level:

```bash
python scripts/init_context.py
```

This creates:
- `PROJECT.md` - Your project context (goals, constraints, preferences)
- `PERMISSION_CONFIG.yaml` - Your permission boundaries

### 2. Set Your Permission Level

Choose one of three levels:

**Level 1 - Master Mode** (Full autonomy)
- AI makes all technical decisions
- Only confirms: spending money, public messages, deleting databases
- Best for: High trust, high risk tolerance

**Level 2 - Collaborative Mode** (Balanced, recommended)
- AI executes most tasks autonomously
- Confirms: money, public messages, important deletions, system changes
- Best for: Most users, balanced control

**Level 3 - Assistant Mode** (High control)
- AI provides suggestions and code
- Confirms: All operations before execution
- Best for: New users, low risk tolerance, learning mode

### 3. Start with Requirements

Instead of detailed specifications, start with what you want:

```
"I need a team chat tool"
```

AI will ask clarifying questions:
- Who is this for?
- Wh...