power-apps-code-apps

Power Apps Code Apps — Multi-Agent Skill

An orchestrator skill for building, connecting, deploying, and managing Power Apps Code Apps (React/TypeScript on Power Platform). Delegates to 6 specialized agents based on the task domain.

When to Use

• Building or scaffolding a new code app from scratch
• Connecting to data sources (Dataverse, Azure SQL, SharePoint, Copilot Studio, custom APIs)
• Deploying via PAC CLI or npm CLI, setting up ALM pipelines
• Configuring security (CSP), telemetry (App Insights), or embedding (iframe)
• Planning architecture, writing requirements, or managing a code app product roadmap
• Migrating from v0.x to v1.0 of the client library

Agent Routing

Analyze the user request and delegate to the most appropriate agent. If the request spans multiple domains, invoke agents sequentially.

Domain	Agent	Trigger Keywords
Frontend	pa-frontend	React, TypeScript, Vite, scaffold, UI, component, Fluent UI, template, npm run dev, local play, migrate v1, getContext, embed iframe
Backend/Data	pa-backend	data source, Dataverse, Azure SQL, SharePoint, Copilot Studio, CRUD, OData, filter, stored procedure, connector, getAll, create, update, delete, metadata, getMetadata
Infrastructure	pa-infra	deploy, push, ALM, solution, pipeline, CSP, content security policy, App Insights, telemetry, Azure Functions, API Management, custom connector, connection reference
Business Analysis	pa-business-analyst	requirements, user story, acceptance criteria, data model, schema, validation rules, field mapping, entity relationship
Architecture	pa-architect	architecture, design, topology, integration pattern, solution structure, multi-environment, data flow, security model, scalability
Product Management	pa-product-manager	roadmap, feature plan, migration plan, sprint, backlog, release, MVP, prioritization, stakeholder

Procedure

1. Parse the request: Identify which domain(s) the user's question falls into using the routing table above.
2. Load reference knowledge: Read the relevant reference doc(s) from ./references/ to ground the agent's response:
   • Frontend Reference — scaffolding, CLI, migration, context, embedding
   • Backend/Data Reference — all data source connections, CRUD patterns, metadata
   • Infrastructure Reference — ALM, CSP, telemetry, Azure Functions, deployment
3. Delegate to the agent: Invoke the appropriate pa-* agent with the user's request plus relevant context from the reference docs.
4. Synthesize: If multiple agents were consulted, combine their outputs into a unified answer.

Cross-Domain Examples

"Build a code app that reads from Dataverse and displays in a Fluent UI table" → pa-backend (Dataverse connection + CRUD) then pa-frontend (Fluent UI DataGrid component)

"Set up CI/CD for my code app with connection references" → pa-infra (ALM pipelines + connection references)

"I need requirements for an asset management code app with Azure SQL backend" → pa-business-analyst (requirements) then pa-architect (data source topology) then pa-backend (Azure SQL setup)

"Plan migration from v0.3 to v1 and deploy to production" → pa-frontend (v1 migration steps) then pa-infra (deployment workflow)

Platform Compatibility

This skill works across:

• Copilot VS Code Chat: Agents in .github/agents/, skill in .agents/skills/
• Copilot CLI: Same file structure, agents discovered by description
• Claude Code / Claude Code CLI: Copy to .claude/skills/ using the install script

See README.md for installation instructions.