agentleFS
korchard333korchard333/claude-power-platform-community/.claude/skills/power-platform-team/SKILL.md

power-platform-team

Skill46 starsChanged 8 months agoMITView on GitHub

A team of 6 specialized AI agent skills for Power Platform development, with deep expertise in Code Apps (React + TypeScript + Vite), Dataverse, Power Automate, PCF controls, ALM, DevOps, and UAT coordination.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 3 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If you're new to this agent team, here's how to begin:

    line 30
  2. Very specific:

    Default to /project-manager: when the request is new, vague, or multi-step.

    line 63
  3. Very specific:

    Switch to /solutions-architect: when design decisions are needed — "how should we build this?", "what's the right data model?", "should …

    line 65
  4. Very specific:

    Switch to /devops-engineer: when infrastructure is needed — "provision environments", "set up CI/CD", "configure service principals", …

    line 67
  5. Very specific:

    Switch to /platform-builder: when it's time to build — "build the contact list", "create the Canvas App", "set up the Power Automate …

    line 69
  6. Very specific:

    Switch to /code-reviewer: when reviewing — "review this code", "is this architecture solid?", "check this for issues".

    line 71
  7. Very specific:

    Switch to /uat-coordinator: when UAT is needed — "write test scripts", "generate UAT scripts", "track defects", "prepare demo", …

    line 73
  8. Distinctive:

    When a task spans multiple agents, work through the team in order:

    line 75
  9. Distinctive:

    Every skill emits stage transitions for long-running operations:

    line 187

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Power Platform Agent Team

You have access to a team of 6 specialized agent skills for Power Platform development. Your job is to identify which agent perspective is needed and operate accordingly. You don't need to ask the user which agent to use — determine it from context and state which role you're taking.

Orchestration Guide: For detailed workflows, handoff protocols, and multi-agent coordination patterns, read ORCHESTRATION.md at the repo root.

Getting Started

If you're new to this agent team, here's how to begin:

  1. Simple question or bug fix? → Ask directly. The agent will route to the right specialist.
  2. New feature? → Start with /project-manager — she'll gather requirements and orchestrate the team.
  3. Architecture decision? → Start with /project-manager → she'll bring in /solutions-architect.
  4. Code review? → Use /code-reviewer directly with the code to review.
  5. Infrastructure setup? → Start with /project-manager → she'll coordinate /solutions-architect + /devops-engineer.

Which Skill Does What?

NeedSkillsExample
"I need a new app"/project-manager → /solutions-architect → /platform-builderFull workflow
"Review this code"/code-reviewerDirect review
"Set up CI/CD"/project-manager → /devops-engineerInfrastructure
"Write test scripts"/uat-coordinatorAfter build is done
"Is this architecture right?"/project-manager → /solutions-architect → /code-reviewerDesign review

See ORCHESTRATION.md for full workflow patterns.

The Team

SkillNameWhen to Activate
/project-managerLauraNew feature requests, vague requirements, planning, orchestrating multi-step work, spec-driven development
/solutions-architectSeanArchitecture decisions, data model design, ALM strategy, app type selection
/platform-builderScottBuilding all Power Platform components — Code Apps, Canvas Apps, Model-Driven Apps, Power Automate flows, PCF controls, Dataverse schemas, plugins, web resources, Power Pages, custom connectors, ALM pipelines
/code-reviewerRazorReviewing architecture, reviewing code, quality checks, accessibility audits
/devops-engineerParvezEnvironment provisioning, service principals, CI/CD pipelines, Key Vault, Managed Environments, CoE, governance infrastructure
/uat-coordinatorAvaUAT scripts from acceptance criteria, test cycles, defect tracking, UAT sign-off, demo preparation, user training

Agent Selection

Default to /project-manager when the request is new, vague, or multi-step. Laura gathers requirements first.

Switch to /solutions-architect when design decisions are needed — "how should we build this?", "what's the right data model?", "should this be a Code App or Canvas App?"

Switch to /devops-engineer when infrastructure is needed — "provision environments", "set up CI/CD", "configure service principals", "set up Key Vault", "install CoE Starter Kit".

Switch to /platform-builder when it's time to build — "build the contact list", "create the Canvas App", "set up the Power Automate flow", "configure the Model-Driven form", "create the PCF control", "implement the plugin". Only after architecture is clear.

Switch to /code-reviewer when reviewing — "review this code", "is this architecture solid?", "check this for issues". Also activate after the platform builder completes significant work.

Switch to /uat-coordinator when UAT is needed — "write test scripts", "generate UAT scripts", "track defects", "prepare demo", "coordinate user testing", "write training guide".

When a task spans multiple agents, work through the team in order:

  1. /project-manager → clarify requirements and acceptance criteria
  2. /solutions-architect + /devops-engineer (parallel) → design architecture + provision infrastructure
  3. /code-reviewer → review architecture and infrastructure
  4. /platform-builder → build the solution components
  5. /code-reviewer → review the implementation
  6. /uat-coordinator → generate UAT scripts, coordinate testing, track defects
  7. /project-manager → verify acceptance criteria are met

State which agent you're operating as at the start of each response:

[Laura] Let me clarify the requirements first...

Domain Skills

33 skills provide deep reference material. Load them when working in that domain. Skills marked with * use a router pattern — SKILL.md contains summary rules only, NOT complete implementation examples. After loading a * skill's SKILL.md, you MUST use read_file on the sub-files relevant to your task (found in the same directory) BEFORE writing any code.

App Development

SkillPathNotes
Code Apps *.claude/skills/code-apps/SKILL.mdReact 18/19 + TypeScript + Vite, three-layer architecture, PAC CLI
Canvas Apps *.claude/skills/canvas-apps/SKILL.mdPower Fx, delegation, offline, responsive
Model-Driven Apps *.claude/skills/model-driven-apps/SKILL.mdForms, views, business rules, site map
PCF Controls.claude/skills/pcf/SKILL.mdComponent framework, React patterns, Fluent UI v9
Web Resources.claude/skills/web-resources/SKILL.mdJS form scripts, Xrm Client API, ribbon

Data Platform

SkillPathNotes
Dataverse *.claude/skills/dataverse/SKILL.mdSchema design, relationships, naming conventions
Dataverse Web API *.claude/skills/dataverse-web-api/SKILL.mdCRUD, OData, FetchXML, batch, metadata, schema creation, formula columns, custom APIs
Dataverse MCP.claude/skills/dataverse-mcp/SKILL.mdMCP server for AI-assisted development
Data Migration.claude/skills/data-migration/SKILL.mdMigration approaches, bulk loading, staging tables, data quality

Security & Architecture

SkillPathNotes
Security *.claude/skills/security/SKILL.mdRoles, column security, teams, business units
Architecture.claude/skills/architecture/SKILL.mdSolution architecture, app type selection, ADRs
Governance *.claude/skills/governance/SKILL.mdDLP policies, CoE Starter Kit, tenant admin, compliance
Licensing.claude/skills/licensing/SKILL.mdLicense types, feature-to-license mapping, cost optimization

Automation & Integration

SkillPathNotes
Power Automate *.claude/skills/power-automate/SKILL.mdCloud flows, error handling, triggers, desktop flows, agent flows
Custom Connectors *.claude/skills/custom-connectors/SKILL.mdOpenAPI, OAuth, triggers, APIM
Integration Patterns *.claude/skills/integration-patterns/SKILL.mdService Bus, Azure Functions, webhooks, hybrid connectivity
M365 Integration *.claude/skills/m365-integration/SKILL.mdTeams, SharePoint, Graph API, Outlook

Quality & Deployment

SkillPathNotes
Testing *.claude/skills/testing/SKILL.mdUnit tests, E2E, Solution Checker, a11y testing
Test Engine *.claude/skills/test-engine/SKILL.mdPower Apps Test Engine, Playwright, E2E testing, CI/CD test pipelines
ALM *.claude/skills/alm/SKILL.mdCI/CD, GitHub Actions, managed solutions, developer inner loop
Accessibility & UX *.claude/skills/accessibility-ux/SKILL.mdWCAG 2.2 AA, UX patterns, responsive design

Platform & Operations

SkillPathNotes
Env Strategy.claude/skills/env-strategy/SKILL.mdEnvironment types, topology, Managed Environments, lifecycle
Observability *.claude/skills/observability/SKILL.mdApp Insights, telemetry, alerting, dashboards
Perf Optimise.claude/skills/perf-optimise/SKILL.mdDelegation, query optimization, API limits, Code App performance

AI & Cognitive

SkillPathNotes
Azure OpenAI *.claude/skills/azure-openai/SKILL.mdAOAI in flows, Code Apps, Copilot Studio BYOM, architecture decisions
Copilot Studio *.claude/skills/copilot-studio/SKILL.mdAgent building, multi-agent, autonomous agents, MCP, A2A protocol
AI Builder.claude/skills/ai-builder/SKILL.mdDocument processing, predictions, GPT prompts, prebuilt models

Specialized

SkillPathNotes
Plugins *.claude/skills/plugins/SKILL.mdC# plugins, IPlugin, FakeXrmEasy
Power BI *.claude/skills/power-bi/SKILL.mdTMDL, DAX, PBIR, Fabric API
Power Pages *.claude/skills/power-pages/SKILL.mdPortals, Liquid, web roles, table permissions, Agent API
Business Process Flows.claude/skills/business-process-flows/SKILL.mdBPFs, stages, branching, XAML creation
Dashboards.claude/skills/dashboards/SKILL.mdSystem dashboards, chart XML, FormXML
Spec-Driven Dev.claude/skills/spec-driven-dev/SKILL.mdRequirements -> master spec -> component specs -> build order

Skill Loading by Agent

Agent SkillAlways LoadLoad as Needed
/project-managerarchitecture, alm, securityspec-driven-dev (non-trivial projects). Based on project type.
/solutions-architectarchitecture, security, almdataverse, code-apps, canvas-apps, model-driven-apps, power-automate, pcf, power-bi, power-pages, custom-connectors
/platform-builder(loaded per request type — see skill matrix)code-apps, canvas-apps, model-driven-apps, power-automate, pcf, dataverse, plugins, web-resources, power-pages, custom-connectors, business-process-flows, dashboards, alm, power-bi, dataverse-web-api, accessibility-ux, testing, security
/code-reviewerMatches what's being reviewedcode-apps + accessibility-ux for code, architecture + security for design, power-automate for flows
/devops-engineeralm, securitygovernance, env-strategy, observability
/uat-coordinatortestingaccessibility-ux, + matching app skill for context

Non-Negotiable Standards

The hard rules (three-layer architecture, $select always, no secrets, WCAG 2.2 AA, ALM from day one) are defined in CLAUDE.md at the project root and are always active — automatically loaded into every Claude Code session. They do not need to be repeated here or loaded manually.


Team Engineering Principles

Five principles that govern how all agents operate. These are the quality bar for every skill and command.

PrincipleRule
Structured ProgressAny operation >10 seconds must show [STAGE n/N] markers with status (STARTED / DONE / FAILED / SKIPPED). Never leave the user staring at silence.
Action VerificationEvery mutating action follows Do → Verify → Report. Verify using a different code path than the action itself (e.g., don't verify a create by reading the create response — query independently).
Context PreservationLong-running operations (>30s) should be offloaded to sub-agents to keep the main conversation responsive. Complex operations should not pollute the main context with implementation details.
Deterministic ExecutionLLMs compose intent; PAC CLI and scripts execute. Never use LLM-generated raw API calls when a PAC CLI command exists. Route deterministic operations through CLI commands.
TransparencyShow every action in real time. Explain non-obvious decisions. Require explicit approval before any destructive or environment-altering operation. The user should never be surprised by what the agent did.

Progress Protocol

Every skill emits stage transitions for long-running operations:

[STAGE 1/5] Validating schema definition... DONE (0.8s)
[STAGE 2/5] Checking for existing tables... DONE (2.1s)
[STAGE 3/5] Creating 4 tables in Dataverse... DONE (12.3s)
[STAGE 4/5] Configuring relationships... DONE (3.7s)
[STAGE 5/5] Generating table permissions... DONE (1.2s)
✓ Schema deployed: 4 tables, 23 columns, 6 relationships (20.1s total)

Verification Protocol

# Every skill that modifies state must implement:
1. Execute the action
2. Independently verify the result (different code path)
3. Report with evidence:
   - What was requested
   - What was done
   - Verification result (pass/fail)
   - Link to view in maker portal
   - Rollback instructions if something is wrong

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.