A framework-agnostic software engineering standard for AI agents.
Engineering Excellence is a reusable AI skill that teaches coding agents to think and act like experienced software engineers instead of code generators.
Rather than focusing on producing code as quickly as possible, this skill enforces a disciplined engineering workflow that prioritizes quality, maintainability, correctness, security, testing, and long-term sustainability.
It is designed to work with any programming language, any framework, and any technology stack.
- 📋 Spec-Driven Development (SDD)
- 🧪 Test-Driven Development (TDD)
- 🏗️ SOLID Architecture
- 🧩 Modular Design
- 🧹 Clean Code
- 🔄 DRY
- 🎯 KISS
- 🚫 YAGNI
- 🔒 Security by Design
- ⚡ Performance by Design
- ♿ Accessibility by Design
- 🌐 Internationalization (i18n)
- 🔎 SEO & Agentic Discoverability
- 🧭 Environment-aware behavior
- ✅ W3C Compliance
- 🚀 Production Readiness
- 📚 Documentation-first workflow
- 🐳 Docker best practices
- 📦 Dependency management
- 🔍 Code review guidelines
- 🚦 Quality Gates
- 📊 Engineering Score
Writing code is easy.
Building software that is maintainable, testable, secure, scalable and understandable years later is much harder.
Engineering Excellence exists to help AI agents consistently make engineering decisions that experienced software developers would make.
The goal is not simply to generate working code.
The goal is to engineer software that can confidently reach production.
Engineering Excellence is intentionally independent of any language or framework.
It works equally well with:
- TypeScript
- JavaScript
- Python
- Go
- Rust
- Java
- Kotlin
- C#
- PHP
- Ruby
and frameworks such as:
- Nuxt
- Vue
- React
- Angular
- Laravel
- Symfony
- Spring Boot
- ASP.NET
- FastAPI
- Express
- NestJS
or any other technology stack.
This skill encourages AI agents to consistently apply:
- SOLID
- DRY
- KISS
- YAGNI
- Separation of Concerns
- Composition over Inheritance
- High Cohesion
- Loose Coupling
- Clean Architecture (when appropriate)
Every implementation follows the same engineering process:
- Understand requirements
- Produce or update the specification
- Ask questions if information is missing
- Wait for approval
- Write failing tests first (TDD)
- Implement the smallest possible solution
- Refactor
- Execute quality gates
- Update documentation
- Produce an Engineering Report
Testing is considered mandatory.
Recommended minimum quality targets:
| Metric | Target |
|---|---|
| Line Coverage | 95% |
| Function Coverage | 95% |
| Statement Coverage | 95% |
| Branch Coverage | 90% |
A task is not considered complete while tests are failing.
The skill promotes secure software development by encouraging:
- Input validation
- Secure defaults
- Secret management
- Principle of least privilege
- Safe dependency usage
- Defensive programming
Performance is treated as a functional requirement.
The skill encourages AI agents to:
- Measure before optimizing
- Optimize Core Web Vitals and Lighthouse scores
- Minimize allocations
- Reduce unnecessary rendering
- Optimize bundle size and prefer lazy loading when beneficial
- Eliminate performance bottlenecks
It aims for the highest realistically achievable scores — never sacrificing maintainability just to move a benchmark.
Accessibility is a first-class requirement, never traded away for aesthetics. Whenever applicable, the generated software should target:
- WCAG 2.2 AA
- Semantic HTML
- Keyboard navigation
- Accessible forms
- Proper focus management
- ARIA only when necessary
- W3C-compliant HTML and CSS
Whenever applicable, applications are designed with internationalization in mind:
- No hardcoded user-facing strings
- Copy separated from logic
- An architecture that allows future localization without major refactoring
For public-facing applications, the skill generates the artifacts that make content discoverable by both search engines and AI agents:
- Metadata, canonical URLs, and structured metadata
sitemap.xmlandrobots.txtllms.txt/llms-full.txtfor AI consumption- Semantic HTML and no duplicate content
Behavior adapts automatically to the deployment environment:
- Development and staging →
noindex,nofollow - Production → indexing enabled
- Administration panels, dashboards, and authenticated areas are never indexed
Production-ready software, when applicable, ships with i18n, accessibility, SEO and
agentic SEO, robots.txt / sitemap.xml / llms.txt, security headers and CSP,
compression, caching, health endpoints, structured logging, monitoring hooks, Docker,
CI/CD, and environment-specific configuration.
Engineering Excellence recommends reproducible development environments.
Docker should be used whenever it improves consistency and developer experience, while avoiding unnecessary complexity for simple projects.
Every dependency should have a clear justification.
The skill encourages:
- Native APIs whenever practical
- Well-maintained libraries
- Minimal dependency footprint
- Regular updates
- Avoiding unnecessary vendor lock-in
Before completing any task, the following checks should pass:
- ✅ Specification updated
- ✅ Tests passing
- ✅ Coverage requirements satisfied
- ✅ Lint passing
- ✅ Build successful
- ✅ Documentation updated
- ✅ No duplicated code
- ✅ No dead code
- ✅ Security reviewed
- ✅ Performance reviewed
- ✅ Accessibility reviewed
At the end of every implementation, AI agents are encouraged to produce a final report summarizing:
- Specification
- Architecture
- Testing
- Coverage
- Security
- Performance
- Accessibility
- Maintainability
- Technical Debt
- Overall Engineering Score
Contributions are welcome.
Ideas, improvements, corrections and new engineering guidelines are encouraged through Issues and Pull Requests.
MIT License.
Build software. Engineer excellence.