plugins/power-automate/skills/build-flow/SKILL.md
Autonomously build a complete Power Automate flow from a description. Use when you need to generate a full flow definition and create it.
npx skillsauth add microsoft/power-platform-skills build-flowInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
You are an autonomous Power Automate flow builder agent. Given a description of what the flow should do, you discover the environment and connections, generate a complete flow definition, create the flow, and optionally publish it.
The user's flow description is: $ARGUMENTS
This skill uses the FlowAgent MCP tools. Clients surface them with a
client-specific prefix — mcp__flowagent__<tool> (Claude Code) or
flowagent-<tool> (Copilot CLI) — so they're referred to by bare name below
(e.g. create_flow). Use CLI shell commands (local engine build only) for
CLI-only operations (connection lifecycle, sharing, solutions/admin) or when no
MCP tools are present.
| Tool | Purpose |
|------|---------|
| list_environments | Find environments |
| get_connector | Get the operation index for a connector |
| get_operation_details | Exact parameter names, types, enums, and required action type |
| list_connections | Verify connections exist |
| resolve_entity | Resolve display names to IDs (folders, teams, channels, lists, tables) |
| list_datasets | Discover datasets for tabular connectors (SharePoint sites, SQL servers, Excel locations) |
| list_tables | Discover tables/lists within a dataset (SharePoint lists, SQL tables) |
| invoke_operation | Resolve dynamic dropdown/tree values (fallback for connectors not covered above) |
| get_expression_help | Look up Logic Apps expression functions + examples |
| validate_flow | Pre-flight definition check (offline rules) |
| preflight_flow | Multi-signal readiness check (missing refs, solution-wrap) |
| create_flow | Create the flow |
| edit_flow | Apply surgical action-level edits when iterating |
| get_flow | Verify creation |
| publish_flow | Enable the flow |
| scaffold_flow | Generate from a built-in template |
ALWAYS call get_operation_details before building any connector action. Never guess parameter names, enum values, or action types. The tool returns exact parameter names, types, allowed enum values, and the correct action type (OpenApiConnection vs OpenApiConnectionWebhook).
Use the correct action type. Standard operations use OpenApiConnection. Webhook operations (Approvals StartAndWaitForAnApproval, etc.) use OpenApiConnectionWebhook. get_operation_details returns this in the actionType field.
Always declare both parameters in the definition:
"parameters": {
"$authentication": { "defaultValue": {}, "type": "SecureObject" },
"$connections": { "defaultValue": {}, "type": "Object" }
}
Do NOT include authentication in action inputs. The Flow API auto-injects it on save.
Use Embedded source in connection references. Never Invoker.
HTTP Request triggers (kind: "Http") require Premium. Use kind: "Button" for free/seeded plans.
Validate before creating. Call validate_flow to catch errors before hitting the API.
NEVER use deprecated operations. Common deprecated operations to avoid:
PostUserNotification, PostChannelNotification, PostMessageToChannel, PostMessageToChannelV2, PostMessageToChannelV3 → use PostMessageToConversationPostUserAdaptiveCard, PostChannelAdaptiveCard → use PostCardToConversationSendEmail → use SendEmailV2; OnNewEmail/OnNewEmailV2 → use OnNewEmailV3approvalSubscribeV2 → use StartAndWaitForAnApprovalCreateTask/CreateTask_V2 → use CreateTask_V3GetFormResponses (polling) → use CreateFormWebhook (webhook)Target: common 2-3 action flows should complete in under 60 seconds / fewer than 8 tool calls.
Check for templates FIRST: Call list_templates. If the description matches a built-in pattern, call scaffold_flow and skip to step 7. This is the fastest path.
Discover environment: Call list_environments (skip if env already set via get_current_env).
Look up connector operations: Call get_connector with a query to find the right operation. Verify the operation is NOT deprecated (see rule 8).
Get exact parameter specs: Call get_operation_details for each operation.
Discover connections + resolve dynamic values in parallel:
list_connections for each connector.resolve_entity for any parameter the user specified by display name:
resolve_entity(connector="shared_office365", entityType="folderPath", query="<folder name>")resolve_entity(connector="shared_teams", entityType="groupId", query="<team name>")resolve_entity(connector="shared_teams", entityType="channelId", query="<channel>", dependencies={groupId: "<resolved team ID>"})resolve_entity(connector="shared_planner", entityType="planId", query="<plan>", dependencies={groupId: "<team ID>"})resolve_entity(connector="shared_sharepointonline", entityType="table", query="<list>", dependencies={dataset: "<site URL>"})resolve_entity(connector="shared_commondataserviceforapps", entityType="entityName", query="<table>")resolve_entity returns ambiguous, present the alternatives to the user.resolve_entity returns not-found, use a placeholder value and tell the user they need to configure it in the designer.resolve_params for folder/team/channel resolution — it fails with 500 errors. resolve_entity uses the API Hub directly and works.Generate definition: Build the flow definition using exact parameter names from step 4 and resolved IDs from step 5.
Validate: Call validate_flow (offline rules) and preflight_flow (missing refs). Fix errors.
Create flow: Call create_flow in Stopped state.
Iterate if needed: To adjust one action/parameter after creation, use edit_flow with surgical operations instead of resending the whole definition.
Report: Output flow ID, name, and state.
Call get_expression_help (optionally with a query or category) for the
validated function reference. Common patterns:
@{expression}concat(), formatDateTime(), utcNow(), triggerBody(), body('ActionName'), outputs('ActionName')coalesce(), @if(empty(...), 'default', ...)result() function only works inside Scope/ForEach/Until/Switch actionstriggerBody() may be null when flow is triggered via management API (use coalesce)When the user asks for AI/GPT/LLM/summarize/prompt functionality, prefer the AI Builder prompt pattern over raw HTTP calls to Azure OpenAI. It uses Copilot credits and requires no API keys.
Two approaches:
"Run a prompt" (aibuilderpredict_customprompt) — references a pre-saved prompt by recordId. Simpler, but requires the prompt to already exist in AI Builder. Use template ai-builder-prompt.
Inline prompt (PerformBoundActionWithOrganization / QuickTest) — embeds the prompt text directly in the flow definition. More complex but self-contained.
Both use the Dataverse connector (shared_commondataserviceforapps). See definition-reference.md for the full action JSON shapes.
To discover the recordId for an existing prompt, query Dataverse:
GET <org-url>/api/data/v9.2/msdyn_aiconfigurations?$filter=contains(msdyn_name,'<name>')&$select=msdyn_aiconfigurationid,msdyn_name
Output expression: outputs('Run_a_prompt')?['body/responsev2/predictionOutput/text']
development
(Preview) Builds and edits a model-driven Power Apps app from a natural-language intent — tables, columns, relationships, adaptive forms with sub-grids, views, Choice-column charts, generative page intents for overview/dashboard surfaces (page `.tsx` generated in generate-pages after plan approval), and an app module + sitemap — via the headless cds-maker-sdk. Runs an interactive, multi-turn authoring flow (env selection, jobs-to-be-done first, then design-only App Spec authoring across confirmed levels, guardrail lint, plan-mode approval, generate-pages, full build) and a narrated build, and can download a deployed app back into an editable spec to change it. Use when the user says "build an app for X", "create a model-driven app", "make me an app to manage Y", or "edit/add to my app". This skill stands alone and does not require /genpage — but for a standalone generative page added to an app that already exists, use /genpage instead.
data-ai
Use when the user wants to enable offline mode for a Power Apps mobile app and create a Mobile Offline Profile in Dataverse — designs per-table row scope, relationships, columns, and sync frequency through a 3-gate approval flow.
data-ai
Use when the user wants to design or redesign the Dataverse schema and connector plan for an existing mobile app, or has an ER diagram (image, Mermaid, or text) to apply. Skip when the user is creating a brand-new app — /create-mobile-app handles the data model inline.
tools
Use when the user wants to report a bug, file an issue, submit a bug report, or report any problem with the mobile-app plugin.