🚀 This project demonstrates APM CLI capabilities - Skills give agents powers, guardrails keep them compliant.
Enterprise-grade web development with AI capabilities + automatic compliance - Watch form-builder skill create contact forms while compliance-rules ensures GDPR compliance.
📦 Skills + Guardrails working together:
- form-builder - React Hook Form + Zod validation patterns (Skill: what agents CAN DO)
- compliance-rules - GDPR, security, audit requirements (Guardrail: what agents MUST FOLLOW)
- design-guidelines - Accessibility, UI/UX standards (Guardrail)
🎯 The narrative: form-builder gives your agent the power to build forms. compliance-rules ensures those forms are GDPR compliant.
User: "Build a contact form"
┌─ form-builder (Skill) ──────────────────────────┐
│ ✓ React Hook Form for controlled inputs │
│ ✓ Zod schema for validation │
│ ✓ Accessible form patterns │
└─────────────────────────────────────────────────┘
↓ constrained by ↓
┌─ compliance-rules (Guardrail) ──────────────────┐
│ ✓ GDPR consent checkbox required │
│ ✓ Data minimization (only collect what's needed)│
│ ✓ Audit logging for form submissions │
└─────────────────────────────────────────────────┘
Result: AI builds a production-ready, compliant contact form.
Note
📋 APM Setup Required: First install APM CLI from github/apm-cli - follow their setup guide for tokens and runtime configuration.
# 1. Clone and enter the project
git clone https://github.com/github/corporate-website
cd corporate-website
# 2. Install APM dependencies (compliance + design packages)
apm install
# 3. Generate AGENTS.md with enterprise context
apm compile
# 4. Run your first enterprise workflow
# This will trigger hello-world.prompt.md as defined in apm.yml
apm run startThat's it! Your project now has enterprise-grade AI workflows with automatic compliance and design enforcement.
Here's how skills and guardrails compose together:
name: corporate-website
version: 1.0.0
description: Corporate website with form building and compliance
author: Corporate Team
dependencies:
apm:
- danielmeppiel/form-builder # Skill: form building capability
- danielmeppiel/compliance-rules # Guardrail: GDPR, security
- danielmeppiel/design-guidelines # Guardrail: UI/UX standards
scripts:
start: "codex --skip-git-repo-check hello-world.prompt.md"
# Compliance workflows (automatically discovered from dependencies)
audit: "codex --skip-git-repo-check compliance-audit.prompt.md"
gdpr-check: "codex gdpr-assessment.prompt.md"
legal-review: "codex --skip-git-repo-check legal-review.prompt.md"
# Design workflows (automatically discovered from dependencies)
accessibility: "codex --skip-git-repo-check accessibility-audit.prompt.md"
design-review: "codex --skip-git-repo-check design-review.prompt.md"
style-check: "codex style-guide-check.prompt.md"- Enterprise Workflows - Compliance audits, accessibility checks, legal reviews (.prompt.md files)
- Context Enforcement - GDPR rules, design standards automatically applied to AI responses
- Dependency Composition -
apm_modules/with proven enterprise packages from other projects - Universal Compatibility - Works with any coding agent supporting the Agents.md standard (GitHub Copilot, Codex, etc.)
apm run audit # Run GDPR compliance audit
apm run accessibility # Check WCAG 2.1 AA compliance
apm run design-review # Validate design system adherence
apm run legal-review # Legal compliance verification
apm run gdpr-check # Data handling assessmentThe Challenge: AI agents need exactly the right context for their current task, but loading everything creates cognitive overload, while missing key information breaks workflows.
APM's Innovation: Mathematical algorithms that automatically figure out the best places to put your context (standards, rules, guidelines, policies) so AI agents get exactly what they need, when they need it, without information overload.
Think of it like organizing a company handbook. You could put everything in one giant manual that everyone carries around, or you could smartly distribute relevant sections to different departments.
APM does this automatically for your project context:
# See what APM is planning to do
apm compile --verbose --dry-runReal example from this project - APM found 9 different context files (standards, compliance rules, guidelines) and had to decide where to put each one:
🎯 Smart Distribution Results:
• Design standards → Root level (affects 6 different folder types)
• GDPR compliance → Root level (affects 10 different folder types)
• API development → backend/api/ only (affects 1 specific folder)
• Testing strategy → tests/ only (affects 3 test-related folders)
• React components → Root level (affects 3 component folders)
The Core Problem: Every instruction needs to be accessible to files that need it, but agents shouldn't be overwhelmed with irrelevant instructions.
APM's Solution: Uses "distribution scores" to decide placement:
- 0-30% distribution → Place locally (like
backend/**/*.py→ goes inbackend/api/AGENTS.md) - 30-70% distribution → Smart multi-placement (verify coverage, fallback to root if needed)
- 70%+ distribution → Place at root (like
**/*.{py,js,ts,tsx}→ goes in rootAGENTS.md)
Why This Matters:
- ✅ Coverage Guarantee: Every file can access the instructions it needs
- ⚡ Efficiency: Agents see mostly relevant context (49.6% efficiency in this project)
- 🧠 Cognitive Load: No more overwhelming agents with irrelevant standards
Instead of one massive file, APM created 5 targeted context files:
- Root
/AGENTS.md- Design standards, compliance rules, React patterns (broad patterns) backend/api/AGENTS.md- FastAPI security, database patterns (backend-specific)tests/AGENTS.md- Testing strategy, pytest patterns (testing-specific)docs/AGENTS.md- Documentation standards (docs-specific)scripts/deployment/AGENTS.md- DevOps patterns (deployment-specific)
The Result: When an AI agent works on backend/api/auth.py, it automatically inherits:
- Root standards (design + compliance) ← Always relevant
- Backend-specific API patterns ← Highly relevant
- No testing or documentation noise ← Clean context
✅ Works with any coding agent that supports AGENTS.md: GitHub Copilot, Cursor, Claude, Codex, etc.
Architectural Advantage: APM bridges the gap between maintainable governance (modular .instructions.md source files) and universal compatibility (standard AGENTS.md output), with smart optimization ensuring agents get exactly the context they need.
Replace inconsistent compliance with engineered enterprise standards:
❌ Before APM:
- "Add authentication" → unpredictable compliance violations across team
- Manual accessibility audits catch issues in production
- Design system violations slip through code review
✅ With APM Dependencies:
- Shared context + structured workflows → consistent, compliant outcomes
- AI agents know GDPR requirements before they start coding
- Accessibility and design standards enforced automatically
The Power: Your AI agents understand enterprise compliance, accessibility requirements, and design standards from day one.
- Frontend: Vanilla HTML, CSS, JavaScript with Vite build system
- APM Dependencies: Enterprise compliance + design packages
- AI Compatibility: Works with GitHub Copilot, Claude Code, Codex, and other Agents.md compatible tools
- 🔒 GDPR Compliance: Data handling, retention policies, audit trails
- ⚖️ Legal Review: Automated compliance checking workflows
- 📊 Audit Trails: User interaction logging for compliance
- 🛡️ Security Standards: Encryption, authentication requirements
- ♿ Accessibility: WCAG 2.1 AA compliance, screen reader support
- 🎨 Design System: Color palettes, typography scales, component standards
- 📱 Responsive Design: Mobile-first, touch target requirements
- ⚡ Performance: Optimization guidelines, loading strategies
apm run audit # Full enterprise compliance audit
apm run gdpr-check # GDPR assessment and data handling review
apm run legal-review # Legal compliance verificationapm run accessibility # WCAG 2.1 AA compliance check
apm run design-review # Design system adherence validation
apm run style-check # Style guide compliance verificationnpm run dev # Start development server (http://localhost:3000)
npm run build # Build for production
apm compile # Generate AGENTS.md from dependenciesgit clone https://github.com/github/corporate-website
cd corporate-website
apm installnpm install # Install Node.js dependencies
apm compile # Generate context from dependencies- 📖 APM CLI Documentation - Complete APM usage guide
- 🚀 Getting Started with AI-Native Development - Learning path and framework
- 🔧 form-builder Skill - React Hook Form + Zod patterns
- 🏢 compliance-rules Guardrail - GDPR compliance package
- 🎨 design-guidelines Guardrail - Accessibility and design system
- 🤖 Agents.md Standard - Universal AI agent compatibility
Learning Guide — Awesome AI Native
A practical companion guide for AI-Native Development: https://danielmeppiel.github.io/awesome-ai-native
A friendly, step-by-step example-driven learning path for AI-Native Development — leveraging APM CLI patterns.
APM transforms enterprise development with reliable, compliant AI workflows