al-development-collection
Advanced tools
AI Native AL Development toolkit for Microsoft Dynamics 365 Business Central with GitHub Copilot integration
AI Native Development toolkit for Microsoft Dynamics 365 Business Central implementing the AI-Native Instructions Architecture framework with 37 specialized Agent Primitives across 3 systematic layers. Features AL Orchestra multi-agent TDD system that transforms feature requests into code following BC best practices.
v2.8.0 - Workflow Simplification: Direct workflow (User → al-architect → al-conductor)
⚡ 5-Minute Setup → Complete Quick Start Guide
Option 1: VS Code Extension (Recommended)
1. Install from Marketplace: Search "AL Development Collection"
2. Open Command Palette (Ctrl+Shift+P)
3. Run: AL Collection: Install Toolkit to Workspace
Option 2: NPM Package
# Install via npm
npm install github:javiarmesto/AL-Development-Collection-for-GitHub-Copilot
npx al-collection install
# Or use directly with npx
npx al-development-collection install
CLI Commands:
# Install to current or specified directory
npx al-development-collection install [path]
# Update existing installation (merge mode)
npx al-development-collection update
# Validate installation
npx al-development-collection validate
# Get help
npx al-development-collection --help
Features:
Option 3: Clone & Install
For development or customization:
git clone https://github.com/javiarmesto/AL-Development-Collection-for-GitHub-Copilot.git
cd AL-Development-Collection-for-GitHub-Copilot
node install.js install [your-al-project]
Then: Reload VS Code (Ctrl+Shift+P → Developer: Reload Window)
Recommendation: Use VS Code Extension for easiest setup. Use npm CLI for automation. Use Clone for development.
📖 Follow the complete example: Customer Loyalty Points System
Or start directly:
New project?
@workspace use al-initialize
Have a feature to build?
Use al-architect mode
[Describe your requirements]
Need debugging?
Use al-debugger mode
[Describe the issue]
This collection implements the AI Native-Instructions Architecture framework with 3 systematic layers that transform ad-hoc AI usage into reliable, repeatable engineering:
graph TB
Dev[👤 Developer] --> Layer1[Layer 1: Markdown Prompt Engineering]
Layer1 --> |Structured Instructions| Layer2[Layer 2: Agent Primitives]
Layer2 --> |Context Optimization| Layer3[Layer 3: Context Engineering]
Layer2 --> Instructions["📋 Instructions Files (Auto-applied guidelines)"]
Layer2 --> Agents["💬 Agents (Role-based specialists)"]
Layer2 --> Prompts["🎯 Prompts-Agentic Workflows (Task execution)"]
Layer3 --> Modular["Modular Loading (applyTo patterns)"]
Layer3 --> AGENTSMD["AGENTS.md Standard (Universal portability)"]
Instructions --> Copilot[🤖 GitHub Copilot]
Agents --> Copilot
Prompts --> Copilot
style Layer1 fill:#9C27B0
style Layer2 fill:#2196F3
style Layer3 fill:#4CAF50
style Copilot fill:#FF9800
| Layer | Focus | Purpose | AL Implementation |
|---|---|---|---|
| Layer 1 | Markdown Prompt Engineering | Structured instructions using semantic markdown | Headers, lists, links for AI reasoning |
| Layer 2 | Agent Primitives | Configurable tools (Instructions, Modes, Prompts) | 7 Instructions + 11 Agents + 18 Workflows + 1 Guide |
| Layer 3 | Context Engineering | Strategic context management via applyTo patterns | Modular loading, AGENTS.md compilation ready |
| Primitive Type | Count | Activation | Purpose |
|---|---|---|---|
| Instructions | 9 files | Auto via applyTo patterns | Persistent coding rules & standards |
| Agents | 7 files | Use [mode-name] | Role-based strategic/tactical specialists |
| Orchestra System | 4 files | Use al-conductor | Multi-agent TDD orchestration |
| Agentic Workflows | 18 files | @workspace use [name] | Complete task execution processes |
| Total Primitives | 37 tools | — | Complete AI Native toolkit with TDD |
Markdown Prompt Engineering implemented as modular .instructions.md files with applyTo patterns:
Always Active (apply to **/*.al):
Context-Activated (apply based on context):
applyTo: **/*.al)applyTo: **/*.al)applyTo: **/test/**/*.al)Integration & Coordination:
.github/copilot-instructions.md)💡 Context Engineering: Instructions use
applyTofrontmatter for selective loading, optimizing context window usage
Systematic processes as .prompt.md files with validation gates and tool access:
General Workflows - Invoke with @workspace use [name]:
Context & Memory Workflows - AI assistant continuity:
Copilot Workflows - Specialized for AI-powered features:
💡 Agentic Workflows: Prompts orchestrate primitives into complete processes with human validation checkpoints.
Role-based specialists with MCP tool boundaries (like professional licensing):
Strategic Specialists (design & analyze):
Tactical Specialist (implement & execute):
💡 Tool Boundaries: Each mode has explicit CAN/CANNOT lists preventing cross-domain security breaches
Multi-Agent TDD Orchestration adapted from GitHub Copilot Orchestra with AL-specific enhancements:
Main Agent:
Specialized Subagents (called automatically by conductor):
Key Features:
.github/plans/Usage:
Use al-conductor mode
Add email validation to Customer table with these requirements:
- Use regex pattern matching
- Allow empty emails (optional in BC)
- Show user-friendly error messages
Result: Multi-phase plan → TDD implementation → Quality validation → Documentation trail
📖 Complete Orchestra Documentation
💡 When to Use: Complex features (3+ phases), production code requiring quality gates, bug fixes with TDD proof, or learning AL through guided implementation
Centralized documentation system ensuring all agents share context and maintain consistency:
Documentation Location: .github/plans/
Core Documents:
architecture.md - System architecture decisions and patternsspec.md - Functional and technical specificationstest-plan.md - Test strategy and coverage requirementsmemory.md - Session history and decision rationaleAgent Integration:
Orchestra Agents (context-aware collaboration):
Specialist Agents (context-aware design):
<endpoint>-api-design.md documenting contracts and integration<feature>-copilot-ux-design.md covering AI prompts and UXKey Benefits:
Workflow Example:
1. Use al-architect mode → Creates architecture.md
2. @workspace use al-spec.create → Generates spec.md
3. Use al-conductor mode → Reads docs, implements with subagents
4. al-review-subagent → Validates against architecture.md + spec.md
5. Result: Consistent implementation aligned with design
Document Templates Available:
docs/templates/api-design-template.mddocs/templates/copilot-ux-design-template.md💡 Best Practice: Always create architecture.md and spec.md before using al-conductor for medium/high complexity features
NEW: Visual guide for choosing the right workflow based on task complexity:
📖 Complete Development Flow Documentation
This comprehensive guide includes:
💡 Quick Start: Unsure which agent or workflow to use? Check the decision tree in the complete flow guide!
📖 New to the collection? Check the Complete Development Flow Guide for visual decision trees and pattern selection!
Have a feature to implement? The collection uses automatic complexity classification with validation gates:
⚠️ Experimental & Customizable Classification: This routing system is an experimental approach that you can customize by editing framework files (
README.md,agents/index.md,instructions/copilot-instructions.md) in your repository. Complexity is assessed by scope, integration depth, and architectural impact—not by counting AL objects. Modify the criteria, thresholds, and routing paths to match your team's expertise and project requirements.
| Complexity | Domain | Scenario Description | Agent Route | Why This Path |
|---|---|---|---|---|
| 🟢 LOW | 🎯 Standard | Simple field addition, basic validation, single UI change | al-developer | Direct implementation - scope is clear, no design needed |
| 🟢 LOW | 🐛 Bug Fix | Known issue with clear reproduction steps | al-debugger → al-developer | Diagnose root cause first, then implement fix with tests |
| 🟢 LOW | ✅ Test Addition | Adding tests to existing well-structured code | al-tester → al-developer | Design test strategy, then implement test cases |
| 🟡 MEDIUM | 🏗️ Feature | Business logic with data flow, internal integrations | al-conductor | TDD orchestration ensures quality across 2-3 phases |
| 🟡 MEDIUM | 🌐 API | RESTful endpoints, OData pages, internal API exposure | al-api → al-conductor | Design API contract first, then implement with TDD |
| 🟡 MEDIUM | 🤖 Copilot Feature | Copilot capability, PromptDialog, basic AI integration | al-copilot → al-conductor | Design AI UX and prompts, then build with quality gates |
| 🟡 MEDIUM | 🐛 Complex Bug | Intermittent issue requiring systematic investigation | al-debugger → al-conductor | Diagnose with profiling, then fix with comprehensive tests |
| 🔴 HIGH | 🏛️ Architecture | Multi-module feature, new patterns, broad impact | al-architect → al-conductor | Design architecture first, then orchestrate TDD implementation |
| 🔴 HIGH | 🌐 Integration | External APIs, OAuth, Azure services, webhooks | al-api → al-architect → al-conductor | API design → System architecture → Secure TDD implementation |
| 🔴 HIGH | ⚡ Performance | System-wide optimization, architectural bottlenecks | al-architect → al-conductor | Analyze architecture impact, design optimization strategy |
Quick examples:
al-developeral-conductoral-api → al-architect → al-conductorStart here if unsure:
Use al-architect mode
I need to [describe your requirement]
al-architect will analyze requirements, design the solution architecture, and recommend the appropriate workflow (al-developer for simple features, al-conductor for complex TDD implementation).
📊 See complete routing matrix and decision tree
1. @workspace use al-diagnose → Diagnose and debug
2. @workspace use al-performance → Profile (if needed)
3. Fix (auto-guidelines active)
4. Use al-tester mode → Test strategy
1. Use al-architect mode → Design API contract
2. Use al-api mode → Implement REST/OData
3. @workspace use al-permissions → Security
4. Use al-tester mode → API test suite
5. @workspace use al-build → Deploy
📖 Complete API Flow with examples
1. Use al-copilot mode → Design AI capability
2. @workspace use al-copilot-capability → Register
3. @workspace use al-copilot-promptdialog → UI
4. @workspace use al-copilot-test → Validate
📖 Complete Copilot Flow with examples
This collection leverages Model Context Protocol (MCP) servers to enhance GitHub Copilot's capabilities with specialized tools:
| Server | Purpose | Key Tools |
|---|---|---|
| al-symbols-mcp | AL object analysis from compiled .app packages (by Stefan Maron) | al_search_objects, al_get_object_definition, al_find_references, al_search_object_members, al_get_object_summary, al_packages |
| context7 | Up-to-date library documentation retrieval | resolve-library-id, get-library-docs |
| microsoft-docs | Official Microsoft/Azure documentation search | microsoft_docs_search, microsoft_code_sample_search, microsoft_docs_fetch |
MCP servers are configured in .vscode/mcp.json:
{
"servers": {
"al-symbols-mcp": {
"type": "stdio",
"command": "npx",
"args": ["al-mcp-server"]
}
}
}
The MCP tools are integrated into specific agents based on their specialization:
al-symbols-mcp tools assigned to:
Prompts with al-symbols-mcp:
💡 Note: al-symbols-mcp requires compiled
.appfiles in.alpackagesdirectory. Ensure you have downloaded symbols (al_download_symbols) before using these tools.
Validate the collection before contributing:
# Install dependencies
npm install
# Run validation
npm run validate
Expected output:
✅ Collection is fully compliant and ready for contribution!
User: "Add email validation to Customer table"
Classification: 🟢 LOW
- Scope: Limited (single table)
- Integration: None
- Route: al-developer mode
Result: Field + validation implemented in ~5 minutes
User: "Build customer loyalty points system"
Classification: 🟡 MEDIUM
- Scope: Moderate (Customer + Sales + Loyalty)
- Integration: Internal (event subscribers)
- Route: al-architect → al-conductor (TDD Orchestra)
Workflow:
1. al-architect designs architecture (20 min)
2. al-conductor orchestrates implementation:
- Planning subagent researches BC objects
- Implement subagent executes 7 TDD phases
- Review subagent validates quality gates
3. Result: 10 AL objects, 63 tests, 100% passing (90 min)
User: "Integrate with external payment gateway (OAuth + webhooks)"
Classification: 🔴 HIGH
- Scope: Extensive (Sales + Finance + Security + API)
- Integration: External (REST API, OAuth 2.0)
- Route: al-api → al-architect → al-conductor
Workflow:
1. al-api designs API contract (1 hour)
2. al-architect plans system architecture (2 hours)
3. al-conductor implements with TDD (1-2 days)
4. Security review + performance profiling
Found an issue or have a suggestion?
Report Issues
Suggest Improvements
Contribution Guidelines
See CONTRIBUTING.md for detailed guidelines.
AI Native-Instructions Architecture with 37 Agent Primitives across 3 layers:
AL-Development-Collection-for-GitHub-Copilot/
├── .github/ # GitHub integration
│ ├── ISSUE_TEMPLATE/ # Issue templates
│ ├── PULL_REQUEST_TEMPLATE.md # PR template
│ └── copilot-instructions.md # Master coordination (symlink)
├── instructions/ # 📋 Layer 1: Instructions (9 files)
│ ├── index.md # Complete guide
│ ├── copilot-instructions.md # Master integration guide
│ ├── al-guidelines.instructions.md # Master hub (applyTo: **/*.{al,json})
│ ├── al-code-style.instructions.md # Code style (applyTo: **/*.al)
│ ├── al-naming-conventions.instructions.md # Naming rules
│ ├── al-performance.instructions.md # Performance patterns
│ ├── al-error-handling.instructions.md # Error handling
│ ├── al-events.instructions.md # Event-driven patterns
│ └── al-testing.instructions.md # Testing (applyTo: **/test/**/*.al)
├── prompts/ # 🎯 Layer 2: Agentic Workflows (18 files)
│ ├── index.md # Complete workflows guide
│ ├── al-initialize.prompt.md # Environment & workspace setup
│ ├── al-diagnose.prompt.md # Debug & troubleshoot
│ ├── al-build.prompt.md # Build & deployment
│ ├── al-events.prompt.md # Event implementation
│ ├── al-performance.prompt.md # Deep profiling
│ ├── al-performance.triage.prompt.md # Quick analysis
│ ├── al-permissions.prompt.md # Permission generation
│ ├── al-migrate.prompt.md # Version migration
│ ├── al-pages.prompt.md # Page Designer
│ ├── al-spec.create.prompt.md # Specifications
│ ├── al-pr-prepare.prompt.md # Pull request prep
│ ├── al-translate.prompt.md # XLF translation
│ ├── al-context.create.prompt.md # Context file generation
│ ├── al-memory.create.prompt.md # Memory file generation
│ ├── al-copilot-capability.prompt.md # Copilot capability registration
│ ├── al-copilot-promptdialog.prompt.md # PromptDialog creation
│ ├── al-copilot-test.prompt.md # AI Test Toolkit integration
│ └── al-copilot-generate.prompt.md # Natural language to code
├── agents/ # 💬 Layer 2: Agents (6 + 4 files)
│ ├── index.md # Complete agent guide
│ ├── al-architect.agent.md # 🏗️ Solution architecture (START HERE)
│ ├── al-developer.agent.md # 💻 Tactical implementation
│ ├── al-debugger.agent.md # 🐛 Deep debugging
│ ├── al-tester.agent.md # ✅ Testing strategy
│ ├── al-api.agent.md # 🌐 API development
│ ├── al-copilot.agent.md # 🤖 AI features
│ └── orchestration/ # 🎭 Orchestra System (4 files)
│ ├── README.md # Complete Orchestra docs (23k chars)
│ ├── index.md # Quick reference
│ ├── al-conductor.agent.md # Main orchestration agent
│ ├── al-planning-subagent.agent.md # AL-aware research
│ ├── al-implement-subagent.agent.md # TDD implementation (Haiku 4.5)
│ └── al-review-subagent.agent.md # Quality validation
├── docs/ # Documentation mirrors
│ ├── agents/ # Agent documentation
│ ├── instructions/ # Instructions documentation
│ ├── prompts/ # Prompts documentation
│ └── workflows/
│ └── complete-development-flow.md # Visual workflow guide
├── collections/ # Collection manifests
│ └── al-development.collection.yml # Main collection (37 primitives)
├── references/ # Framework documentation
│ ├── AI Native-INSTRUCTIONS-ARCHITECTURE.md # Framework compliance
│ ├── ai native-concepts.md # Core concepts
│ └── ai native structure.md # Structure guide
├── archive/ # Historical files
│ ├── README.md # Archive documentation
│ ├── MIGRATION-FROM-ORCHESTRATOR.md # Migration guide (v2.8.0)
│ └── al-orchestrator.agent.md # Archived: Smart router (deprecated v2.8.0)
├── al-development.md # Framework overview
├── CONTRIBUTING.md # Contribution guidelines
├── CHANGELOG.md # Version history
├── LICENSE # MIT License
├── SECURITY.md # Security policy
├── validate-al-collection.js # Compliance validation
├── package.json # Dependencies
├── QUICK-START.md # Quick start guide with complete example
├── REPRODUCIBLE-EXAMPLE.md # Step-by-step reproducible example
└── README.md # This file
# Total: 37 Agent Primitives (9 instructions + 18 workflows + 6 agents + 4 orchestra)
# AGENTS.md compilation ready (Layer 3: Context Engineering)
# All primitives use applyTo patterns for modular context loading
.github/plans/
.local/test-execution-results-2025-11-09.md.github/plans/ with complete audit trailapplyTo patterns for modular context loadingMIT License - see LICENSE file for details.
javiarmesto
Status: ✅ AI Native-Instructions Architecture Compliant
Framework: AI Native-Instructions Architecture
Last Updated: 2025-11-14
Collection ID: al-development
Version: 2.8.0
Total Primitives: 37 (9 instructions + 18 workflows + 6 agents + 4 orchestra)
Installation: npm CLI (recommended) | Clone & Install
Context Standard: AGENTS.md ready
Latest: Workflow Simplification - Direct User → al-architect → al-conductor path
FAQs
AI Native AL Development toolkit for Microsoft Dynamics 365 Business Central with GitHub Copilot integration
The npm package al-development-collection receives a total of 9 weekly downloads. As such, al-development-collection popularity was classified as not popular.
We found that al-development-collection demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
pnpm 11.5 now recognizes npm staged publish approvals in release metadata, preventing those releases from being mistaken for lower-trust package publishes.
Security News
Federal audit finds NIST lacked a plan to clear the NVD backlog, wasted funds on duplicate work, and delayed use of CISA data.
Research
/Security News
A mini Shai-Hulud campaign compromised Red Hat Cloud Services npm packages to steal developer and CI/CD secrets during installation.