Production-grade Focused skill library

Practical .NET guidance
for coding agents

A focused collection of SKILL.md guidance built to reduce repeated prompting, unnecessary token use, and avoidable rework when developing production .NET software with AI coding tools.

Canonical skills

4 agent hosts

Clean as You Code

Created by Devi Prakash

Browse the skills Create an install command Release notes View on GitHub

Quick install

Build your command

Make three choices. Start with the recommended defaults if you are unsure.

Safe defaults selected

1 Choose an agent Installation target

2 Choose where

My machineAvailable across repositories This projectStored with one repository

3 Choose a mode

CopyRecommended and portable LinkFor local development

Ready to run Paste this into your terminal

Node.js

$npx dotnet-production-agent-skills install --target all

The installer checks for conflicts before writing and never silently replaces user-authored files.

View every CLI command

Skills Catalog

Find only the guidance relevant to the task at hand. Search by technology, workflow, or engineering concern.

Designed for less rework

Project-wide Copilot instructions guide every coding task. Agents select specialized skills when the request calls for them. Command guardrails are reserved for supported hooks.

🎯

Invariant-First Discipline

Implement the requested behavior without mutating unrelated behavior. Task size changes investigation and test depth, not safety standards.

📚

Progressive Context Disclosure

Agents load concise SKILL.md guidance only when it matches the current task, and open detailed references only when needed.

🛡️

Clean as You Code

Repository analyzers, Roslyn warnings, Sonar rules, and .editorconfig guide new and materially changed code without expanding into unrelated legacy cleanup.

🔄

Composable Workflows

Skills compose naturally. A defect flows from bug-investigation → dotnet-implementation → unit-testing → targeted-validation → diff-review.

🔒

State-Tracked Safety

The TypeScript CLI tracks file digests in .dotnet-agent-skills.json. Repeated installs skip unchanged skills; force operations back up locally edited skills.

🌐

Works Across Agents

Canonical skills follow the portable SKILL.md format. The installer adapts their location for Codex, Copilot, Claude Code, and Cursor.

Agent compatibility

One canonical skill library, installed into the discovery location used by each supported agent. Copilot project installs also manage a section of repository instructions.

Agent Host	Native Format	User Scope Location	Project Scope Location	Installation Method
OpenAI Codex	Portable `SKILL.md`	`~/.agents/skills`	`<repo>/.agents/skills`	Copy / Symlink
GitHub Copilot	Agent Skills	`~/.copilot/skills`	`<repo>/.github/skills`	Copy / Symlink
Claude Code	Portable `SKILL.md`	`~/.claude/skills`	`<repo>/.claude/skills`	Copy / Symlink
Cursor	Agent Skills	`~/.cursor/skills` or `.agents`	`<repo>/.cursor/skills`	Copy / Symlink
Google Antigravity	Agent Skills + Hooks	`~/.agents/skills`	`<repo>/.agents/skills`	Copy / Symlink + Hook

Built from experience

Less repetition.
Less token waste.
Less rework.

I created dotnet-production-agent-skills after repeatedly explaining the same production expectations to GitHub Copilot—then spending more tokens correcting and reworking the result. This project captures those lessons as focused, reusable guidance so coding agents can start with better context and produce safer .NET code the first time.

Devi Prakash Creator and maintainer

GitHub profile View repository

Practical .NET guidance for coding agents