Power Automate

Skill: Power Automate Cloud Flows

When to Use

Trigger when designing, building, reviewing, or debugging Power Automate cloud flows.

---

Flow Design Principles

1. Idempotent flows: Re-running the same trigger should produce the same result
2. Error handling first: Add try/catch scope before writing the happy path
3. Minimal permissions: Use service principal connections for production
4. No hardcoded IDs: Use environment variables or dynamic lookups
5. Name every action: Descriptive action names are your documentation
6. Solution-aware: Every production flow must be inside a solution

---

When to Use Power Automate vs Alternatives

| Scenario | Recommended | Why | |---|---|---| | React to Dataverse record changes | Power Automate | Native trigger with filtering columns | | Complex business logic on save | Plugin (C#) | Runs synchronously in transaction pipeline | | Reusable API callable from multiple apps | Custom API | Registered in Dataverse, callable via Web API | | Scheduled batch processing | Power Automate | Recurrence trigger + pagination | | Real-time validation before save | Plugin (C#) | Pre-operation stage, can throw to block save | | External system integration (async) | Power Automate | HTTP connector with retry policies | | External system integration (sync) | Custom Connector | When response needed immediately | | UI automation (no API available) | Desktop Flow | Power Automate Desktop with cloud orchestration |

---

Solution-Aware Flow Checklist

• [ ] Flow is inside a solution (not "My Flows")
• [ ] All connections use Connection References
• [ ] All hardcoded URLs/IDs use Environment Variables
• [ ] Flow has error handling scope (try/catch)
• [ ] Flow name follows convention: [Solution] - [Trigger] - [Action]
• [ ] Run-only permissions configured if shared
• [ ] Trigger conditions set to avoid unnecessary runs
• [ ] Concurrency controls set on Apply to Each loops
• [ ] Action retry policies configured for external calls

---

⚠️ Verify Flow Patterns Against MS Learn

Before writing flow definitions programmatically, verify patterns using the MS Learn MCP tools:

1. microsoft_docs_search("power automate [your pattern]") — quick overview
2. microsoft_docs_fetch(url) — full page detail
3. microsoft_code_sample_search("power automate [pattern]") — working examples

Skill files are battle-tested but can contain errors. If MS docs contradict what's in the skill file, trust MS docs and flag the discrepancy.

---

⚠️ REQUIRED: Load Sub-Files Before Implementation

SKILL.md is a summary only — it is NOT sufficient for implementation.

The detailed content (complete payloads, XML templates, working examples, edge-case handling) lives in sub-files in the same directory as this SKILL.md. Before writing any code, you MUST use read_file on the sub-files relevant to your task:

• [Error Handling](error-handling.md) — Scope-based try/catch/finally pattern, error notification, cleanup actions
• [Triggers](triggers.md) — Trigger selection, Dataverse filtering columns, trigger conditions (OData expressions)
• [Dataverse Actions & Performance](dataverse-actions.md) — Native connector actions, OData filters, Apply to Each concurrency, child flows, pagination pattern, performance profiles
• [Expressions & HTTP Connector](expressions.md) — Null-safe access, date formatting, JSON parsing, HTTP resilient pattern with retry policies
• [Approvals & Teams Adaptive Cards](approvals-adaptive-cards.md) — Approval flow pattern, Teams card posting, email notification sync, approval types
• [Desktop Flows (RPA)](desktop-flows.md) — When to use RPA, execution modes, cloud-to-desktop integration, best practices
• [Agent Flows](agent-flows.md) — Agent Flows (AI-native automation), generative actions, Automation Center governance, autonomous agent patterns
• [Flow Creation via Web API](flow-creation-api.md) — Programmatic flow creation, activation, and monitoring via Dataverse Web API: clientdata structure, trigger types, connection reference wiring, activation latency, run monitoring via flowruns entity

---

Anti-Patterns

• Flows outside of solutions ("My Flows") — can't be promoted via ALM
• Dataverse calls inside unbounded Apply to Each (no concurrency limit)
• Personal connections instead of connection references
• Missing error handling (no try/catch scope)
• No retry policy on HTTP actions
• Hardcoded GUIDs, URLs, or environment-specific values
• Using "Send an HTTP request to Dataverse" when a native action exists
• Apply to Each with >1000 items and no pagination strategy
• Approval flows with Teams cards AND email notifications enabled simultaneously (card won't sync)
• Using desktop flows when a cloud connector exists for the same system

---

Related Skills

• dataverse — Schema design for tables that flows operate on
• code-apps — Triggering flows from Code Apps via Custom API or Dataverse record triggers
• custom-connectors — Building connectors for external APIs used in flows
• alm — Solution packaging, connection references, environment variables for flow deployment
• plugins — Alternative to flows for synchronous, transactional Dataverse logic

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: korchard333
• Source: korchard333/claude-power-platform-community
• License: MIT

Install and usage instructions live in the source repository linked above.