Code Agreements — Development skill for Claude Code
How I want code written, across all my projects.
How to install Code Agreements
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Qythyx/code-agreements and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Code Agreements does
How I want code written, across all my projects. Maintained by a Claude Code skill.
Alternatives in Development
- Skill Creator — Create new skills, modify existing skills, and measure skill performance 94.1k ★
- Zerostack — Lightweight coding agent written in Rust, optimized for memory footprint and performance 1.7k ★
- Vibe What Do I Want 1.7k ★
README
Code Agreements
How I want code written, across all of my projects.
What this is
These files are read before writing code, so that guidance given once doesn't have to be given again. Some of it was set down up front; most of it accumulates as things get settled during development sessions — usually when I correct, simplify, or rename something.
Each entry records what was actually settled, with the reasoning attached, so a future session can read it and skip the argument. It is not a style guide written from first principles.
What to read
Always read `general.md`. It applies to every language, and it holds the design, comment, and process guidance that is easiest to assume you already know.
Then read the file for **every** language the change touches — plural, not one. A change spanning a C# backend and a TypeScript front end needs `csharp.md` and `typescript-react.md` both.
Layout
| File | Holds |
|---|---|
general.md |
Language-agnostic design and process guidance |
csharp.md |
C# / .NET |
typescript-react.md |
TypeScript, React, and the web front end |
Route an entry by where it **applies**, not where it came up: a design principle that surfaced in C# still belongs in `general.md`.
The bar for a new entry
These files are read in full before every coding task, so each rule taxes every future session whether or not it applies. A new entry has to be **unguessable** (a competent implementer wouldn't arrive at it unaided) and **recurrent** (it fires often, not just conceivably). Most candidates fail one or the other. One entry is a normal yield from a session; five means a session's narrative is being converted into rule-shaped prose.
The entries that earn their keep are mostly preferences — arbitrary, unguessable, two lines. Things I _discovered_ whil
Related Skills
Decompose Wbs Epic
Decompose a WBS epic into smaller, independent, vertical user stories (INVEST), each written as a full 'As a r
Breather
Claude Code will work as long as you will — breather gives you a way out. Tracks real presence across all sess
Open Agreements
OpenAgreements: Fill a Legal Template
Agree
Manage the running list of agreements — the canonical registry of confirmed decisions
Playbook Builder Skill
This repository contains two Claude skills: one for generating organization specific playbook skills, and the
Agent Council
A council of independent code reviewers (Codex, Pi, Devin, OpenCodeReview) for Claude Code: one summary of agr
Related Agents
GitOps CI Specialist
Use this agent when you need to commit code to GitHub and want to ensure CI/CD pipeline success. Examples: Con
Code Improvement Reviewer
Use this agent when you want to review code for quality improvements including readability, performance, and b
Code Improvement Agent
Use this agent when you want to analyze and improve existing code for readability, performance, and best pract