Coding

How This Skill Learns

This skill learns your preferences ONLY from:

• Explicit corrections — "Actually, I prefer X over Y"
• Direct statements — "Always use snake_case"
• Repeated requests — You ask for the same thing 2+ times

This skill NEVER:

• Reads your project files to infer preferences
• Observes without your knowledge
• Stores data you haven't explicitly approved

Memory Storage

Preferences stored at ~/coding/memory.md. Created on first use.

~/coding/
├── memory.md      # Active preferences (≤100 lines)
└── history.md     # Archived old preferences

To create: mkdir -p ~/coding

Memory Format

# Coding Memory

## Stack
- python: prefer 3.11+
- js: use TypeScript always

## Style
- naming: snake_case for Python, camelCase for JS
- imports: absolute over relative

## Structure
- tests: same folder as code, not separate /tests

## Never
- var in JavaScript
- print debugging in production

How Preferences Are Added

1. User corrects output → Agent asks: "Should I remember this preference?"
2. User confirms → Agent adds to ~/coding/memory.md
3. User can review → "Show my coding preferences" lists current memory

No preference is stored without explicit user confirmation.

Rules

• Keep each entry ultra-compact (5 words max)
• Confirm before adding any preference
• Check dimensions.md for categories
• Check criteria.md for when to add
• Never exceed 100 lines in memory.md
• Archive old patterns to history.md

On Session Start

1. Load ~/coding/memory.md if exists
2. Apply stored preferences to responses
3. If no file exists, start with no assumptions

Auxiliary Files

File	Purpose
dimensions.md	Categories of preferences to track
criteria.md	When to suggest adding a preference