coding-1-0-3_skill

Overview

This skill provides a compact, local coding-style memory that adapts to your explicit preferences, conventions, and patterns so code output stays consistent. It stores and applies only confirmed preferences in a small on-disk store under ~/coding/. It never infers habits or reads project files without your clear confirmation.

How this skill works

On first use the skill creates a local ~/coding/ folder and loads memory from memory.md if present. When you correct or confirm a preference the agent asks whether to remember it; only confirmed items are saved in ultra-compact, categorized lines. The skill reads and applies those entries to future code output and supports simple queries to show or remove preferences.

When to use it

• You have explicit style or stack preferences you want applied consistently.
• You want the agent to remember naming, formatting, or structural conventions.
• You prefer the agent avoid inferring rules from your projects or code.
• You need a local, privacy-preserving place for coding preferences.

Best practices

• Give explicit corrections (e.g., “I prefer X over Y”) so the agent can prompt to save.
• Keep each stored entry concise (five words max) for clarity and mergeability.
• Group preferences by category: Stack, Style, Structure, Never.
• Limit memory.md to 100 lines; archive old entries to history.md when needed.
• Confirm before saving and use queries to review or forget entries.

Example use cases

• Remember Python runtime preferences like: python: prefer 3.11+.
• Enforce naming conventions such as: naming: snake_case for files.
• Store test placement choices: tests: colocated, not separate folder.
• Record tooling decisions (Stack): prefer poetry over pip-tools.
• Keep a short blacklist under Never for unwanted tools or patterns.