Celia Coder

A skill that teaches AI coding agents to write and review code for long-term maintainability. One set of design principles governs both — the standards applied when reviewing are the same standards followed when writing.

How It Works

An AI agent loads SKILL.md, which instructs it to load 8 design principle files from principles/. The argument determines the mode:

• Write — "add a caching layer to the database module" — the agent explores the codebase, designs with the principles in mind, implements, and self-reviews before presenting the result.
• Review — src/auth/ or main..feature or #42 — the agent analyzes the code against every principle and reports the top issues ranked by complexity impact.

Principles

Principle	Governs
Module Depth	Deep modules with simple interfaces over many shallow ones
Information Hiding	Encapsulating design decisions, avoiding leakage between modules
Abstraction Layers	Layer separation, pass-through elimination, complexity placement
Cohesion & Separation	Together-or-apart decisions, splitting for abstractions not length
Error Handling	Defining errors out of existence, minimizing exception surface
Naming & Obviousness	Precise names, code that reads without effort
Documentation	Documenting what and why, not how
Strategic Design	Investing in design, resisting tactical shortcuts

Structure

SKILL.md             # The skill definition (write + review)
principles/          # Design principles loaded by the skill
├── module-depth.md
├── information-hiding.md
├── abstraction-layers.md
├── cohesion-separation.md
├── error-handling.md
├── naming-obviousness.md
├── documentation.md
└── strategic-design.md

Adding a Principle

1. Create principles/your-principle.md with four sections: Description, Core Principles, Red Flags, Issue Tags.
2. Add a reference line to the Design Principles section in SKILL.md.