name: add-cloud-flow description: >- Integrates Power Automate cloud flows into a Power Pages site. Lists available flows, suggests relevant ones based on intent, identifies scenarios and web roles, creates metadata files, and generates client-side code to call flows. Handles both new flow registration and adding already-registered flows to additional pages without re-creating metadata. Use when the user wants to add, connect, register, or link a Power Automate cloud flow to their site. user-invocable: true allowed-tools: Read, Write, Edit, Bash, Grep, Glob, AskUserQuestion, Skill, Task, TaskCreate, TaskUpdate, TaskList model: opus

Plugin check: Run node "${PLUGIN_ROOT}/scripts/check-version.js" — if it outputs a message, show it to the user before proceeding.

Add Cloud Flow

Connect one or more Power Automate cloud flows to a Power Pages code site, or wire already-registered flows into additional pages/components. For new flows this skill:

• Creates the adx_cloudflowconsumer metadata YAML in .powerpages-site/cloud-flow-consumer/
• Assigns web roles based on the flow's scenario and target audience
• Generates client-side TypeScript/JavaScript service code to trigger the flow with CSRF authentication

For already-registered flows, the skill skips metadata and role creation and goes straight to client-side integration — wiring the existing flow into new UI locations.

Core Principles

• Multiple flows in one run: The user can add several flows at once. All flows are planned together and implemented together before asking for deployment.
• Ask before acting: Present a full HTML plan (all flows, roles, reasoning) before creating any files.
• Web roles drive access: Every flow must have at least one web role. Anonymous Users role is valid but must be confirmed.
• Scenario determines roles: Understand what each flow does and who triggers it before picking roles.
• Use TaskCreate/TaskUpdate: Track all phases upfront before starting any work.

Prerequisites:

• An existing Power Pages code site with .powerpages-site deployed
• PAC CLI authenticated (pac auth who must succeed)
• Azure CLI authenticated (az login --allow-no-subscriptions works whether or not your account has an Azure subscription — Dataverse and Power Platform tokens are AAD-scoped)

Initial request: $ARGUMENTS

Workflow

1. Verify Prerequisites — Locate the project, confirm .powerpages-site exists, inventory web roles and existing flows
2. List Available Flows — Fetch flows from the Power Automate Flow RP API
3. Select Flows & Understand Scenarios — User picks one or more flows; determine scenario and audience for each
4. Determine Web Roles — Propose minimum roles per flow with reasoning
5. Review Plan — Render HTML plan for all flows; get user approval
6. Create Metadata — Write .cloudflowconsumer.yml for each flow; create missing web roles first
7. Client-Side Integration — Generate typed service code to call each flow from the frontend
8. Verify & Summarize — Validate all YAMLs, record skill usage, offer deployment

Phase 2: List Available Flows

Goal: Fetch all Power Automate cloud flows that have a PowerPages trigger

Actions:

Run the list-cloud-flows script (reads environment from pac auth who internally):

node "${PLUGIN_ROOT}/skills/add-cloud-flow/scripts/list-cloud-flows.js"

This calls the Power Automate Flow RP API with the filter properties/definitionSummary/triggers/any(t: t/kind eq 'powerpages'). Results include:

Separate into two lists:

• Unregistered flows: Flows whose id does NOT match any processid from Phase 1.5 — these need full setup (metadata + roles + client-side)
• Already-registered flows: Flows whose id matches a processid from Phase 1.5 — these already have metadata and roles but can be integrated into additional pages/components

Handle errors:

• Exit code 1 with auth message → prompt the user to run az login --allow-no-subscriptions

• Zero flows from API (no flows with a PowerPages trigger exist at all) → tell the user that no Power Automate flows with a PowerPages trigger were found in this environment. Guide them to create a flow first:

  "No Power Automate cloud flows with a PowerPages trigger were found in this environment. To use this skill, you first need to create a flow in Power Automate with the "When a Power Pages flow step is run" trigger, then run this skill again to connect it to your site."

  Stop the workflow — do not continue to Phase 3.

• Zero unregistered flows but already-registered flows exist → do not stop. Continue to Phase 3 with only the already-registered flows available (for additional frontend integration).

Output: List of unregistered flows and list of already-registered flows

Phase 3: Select Flows & Understand Scenarios

Goal: Understand the user's intent, suggest the most relevant flows (including already-registered ones for additional integration), let the user confirm or adjust, and determine the scenario for each selected flow

Actions:

3.1 Analyze User Intent & Suggest Flows

Before showing the full list, analyze the user's initial request ($ARGUMENTS) against both the unregistered flows and the already-registered flows. Match the user's described need to specific flows using display name, description, and scenario inference.

When matching intent, check already-registered flows first — the user may be asking to wire an existing flow into a new page or component, not register a new one.

• If the user described a specific need (e.g., "add automation for the contact form", "connect the email notification flow"): Identify flows whose name or description closely match the request and present them as recommended picks, explaining why each is a good match. Also show the remaining flows as additional options in case the suggestion is wrong.

  Copy & paste — that's itCopy

  Based on your request, I recommend:
    ⭐ 1. Contact Form Submission — Sends an email when a contact form is submitted (Active)
         → Matches your request to add automation for the contact form
  
  Other available flows:
    2. Support Ticket Handler — Creates a case record from user input (Active)
    3. Newsletter Signup — Adds the user's email to a mailing list (Draft)

• If the user's request is generic (e.g., "add a cloud flow", "connect some flows"): Fall back to presenting the full list without recommendations.

  Copy & paste — that's itCopy

  Available flows:
    1. Contact Form Submission — Sends an email when a contact form is submitted (Active)
    2. Support Ticket Handler — Creates a case record from user input (Active)
    3. Newsletter Signup — Adds the user's email to a mailing list (Draft)

Present both categories clearly when both exist:

New flows (not yet registered):
  1. Newsletter Signup — Adds the user's email to a mailing list (Active)

Already registered (available for additional frontend integration):
  2. Contact Form Submission — Already connected, can be wired into more pages
  3. Support Ticket Handler — Already connected, can be wired into more pages

🚦 Gate (plan · add-cloud-flow:3.1.select-flows): Multi-select over discovered + already-registered flows. Drives the rest of Phases 4–7.

Trigger: Phase 2 list-cloud-flows returned at least one flow. Why we ask: Wrong flows get registered (new .cloudflowconsumer.yml files written) or wrong existing flows get re-wired into the frontend. Cancel leaves: Nothing — no YAML or client code written yet.

Use AskUserQuestion:

If more than 10 flows are available, ask the user to type names or numbers (comma-separated for multiple).

3.2 Tag Selected Flows

For each selected flow, tag it based on its source:

3.3 Determine Scenario Per Flow

For each selected flow, identify its scenario from the name and description:

If a flow's scenario is unclear, use AskUserQuestion per flow:

For integration-only flows, the scenario is still needed to guide where and how the flow is wired into the frontend (Phase 7).

3.4 Determine Client-Side Function Name Per Flow

For each flow, derive a camelCase function name for the client-side service:

• Strip special characters and connector words from the display name
• Example: "PowerPages -> Send an email notification (V3)" → sendEmailNotification
• Example: "Contact Form Submission" → submitContactForm

For integration-only flows, check whether a trigger function already exists in the codebase (from a previous run of this skill). If it does, reuse that function name — do not create a duplicate. Only create a new function if one doesn't exist yet.

Output: Selected flows list, each with tag (new or integration-only), scenario, target audience, and client-side function name

Phase 4: Determine Web Roles

Goal: Propose the minimum set of web roles for each new flow based on its scenario

Skip for integration-only flows — these already have web roles assigned in their .cloudflowconsumer.yml. Do not re-propose or modify roles for them.

Actions:

4.1 Analyze Role Requirements Per Flow (New Flows Only)

Do NOT assign all roles by default. Use the minimum set per flow. Different flows in the same batch can have different role sets.

4.2 Match Against Existing Roles

For each proposed role across all flows, check whether it exists in .powerpages-site/web-roles/. Mark roles as existing or proposed/new.

4.3 Compose Per-Role Reasoning

For each role assigned to each flow, write one sentence explaining why:

• Authenticated Users on a profile form: "Only logged-in users can update their profile."
• Anonymous Users on a contact form: "Unauthenticated visitors must be able to submit a contact request."
• Blog Authors on an email flow: "Blog Authors are the primary actors who submit posts triggering this notification."

4.4 Flag Anonymous Roles

If any flow has Anonymous Users assigned, note this — the HTML plan renders a warning banner and the confirm step must explicitly surface it.

Output: Per-flow role sets with reasoning, list of roles that need creating

Phase 5: Review Plan

Goal: Render the HTML plan for all flows and get user approval before writing any files

Actions:

5.1 Build Plan Data

Assemble the plan JSON (kept in memory — not written to disk). Include all selected flows in CLOUD_FLOWS_DATA, distinguishing new from integration-only:

{
  "SITE_NAME": "<site name>",
  "PLAN_TITLE": "Cloud Flow Integration Plan",
  "SUMMARY": "<N new flows + M integration-only flows, scenarios, key role assignments>",
  "WEB_ROLES_DATA": [
    {
      "id": "<uuid or temp-id>",
      "name": "<role name>",
      "desc": "<description>",
      "builtin": false,
      "isNew": true,
      "isAnonymous": false,
      "color": "#0078d4"
    }
  ],
  "CLOUD_FLOWS_DATA": [
    {
      "flowId": "<workflowEntityId>",
      "name": "<flow display name>",
      "displayName": "<flow display name>",
      "description": "<flow description>",
      "state": "Active",
      "tag": "new | integration-only",
      "scenario": "<scenario type>",
      "rationale": "<why this flow is being added or where it will be additionally integrated>",
      "webRoles": [
        { "id": "<role-uuid>", "reasoning": "<why this role>" }
      ]
    }
  ],
  "RATIONALE_DATA": [
    { "icon": "⚡", "title": "<title>", "desc": "<reasoning>" }
  ]
}

For integration-only flows, webRoles should reflect the existing roles from the .cloudflowconsumer.yml (read-only — not being changed). The rationale should describe where the flow will be additionally integrated (e.g., "Wiring existing Contact Form flow into the support page").

5.2 Render HTML Plan

node "${PLUGIN_ROOT}/scripts/render-cloudflow-plan.js" \
  --output "<PROJECT_ROOT>/docs/cloud-flow-plan.html" \
  --data-inline '<json>'

The render script refuses to overwrite existing files. Before calling it, check if the default output path (<PROJECT_ROOT>/docs/cloud-flow-plan.html) already exists. If it does, choose a new descriptive filename based on context — e.g., cloud-flow-plan-contact-form.html, cloud-flow-plan-apr-2026.html. Pass the chosen name via --output.

Open the rendered file in the default browser (open on macOS, start on Windows, xdg-open on Linux).

5.3 Confirm with User

Give a brief CLI summary: number of flows, scenarios, role count, any anonymous-role warnings.

🚦 Gate (plan · add-cloud-flow:5.3.plan-approval): Final sign-off on the rendered HTML plan before any web role / .cloudflowconsumer.yml / client code is written.

Trigger: Phase 5.2 rendered the HTML plan. Why we ask: Wrong web role assignments committed (especially Anonymous Users on auth-protected flows); orphaned YAML files in .powerpages-site/cloud-flow-consumer/. Cancel leaves: Nothing — no YAML or frontend changes yet.

Use AskUserQuestion:

If "Request changes": Ask what to change, update, re-render, and present again. If "Cancel": Stop.

Output: User-approved plan for all flows

Phase 6: Create Metadata

Goal: Create web roles (if any are new) then write a .cloudflowconsumer.yml for every new flow in the approved plan

Skip for integration-only flows — these already have .cloudflowconsumer.yml and web roles. Do not recreate or modify their metadata. They will be handled in Phase 7 (client-side integration only).

If the approved plan contains only integration-only flows (no new flows), skip this entire phase and proceed to Phase 7.

Actions:

6.1 Create Missing Web Roles

If the approved plan includes roles that do not yet exist (for new flows), invoke /create-webroles now.

After it completes, re-read .powerpages-site/web-roles/ to collect id values for all new roles.

6.2 Process Each New Flow

Repeat steps 6.3–6.4 for every new flow in the approved plan.

6.3 Determine Flow API URL

/_api/cloudflow/v1.0/trigger/<workflowEntityId>

6.4 Run the Create Metadata Script

node "${PLUGIN_ROOT}/skills/add-cloud-flow/scripts/create-cloud-flow-metadata.js" \
  --projectRoot "<PROJECT_ROOT>" \
  --fileSlug "<url-safe-slug>" \
  --flowName "<flow-display-name>" \
  --flowId "<workflowEntityId>" \
  --flowTriggerUrl "" \
  --flowApiUrl "/_api/cloudflow/v1.0/trigger/<workflowEntityId>" \
  --webRoleIds "<uuid1>,<uuid2>" \
  --metadata "<flow-display-name>"

--fileSlug: lowercase display name, special characters replaced with hyphens, max 50 chars. --flowName: display name written as-is into the name YAML field.

The generated YAML (PAC CLI code-site git format):

• adx_* scalar prefix stripped: adx_flowapiurl → flowapiurl, adx_processid → processid
• adx_cloudflowconsumerid → id; adx_websiteid omitted
• M2M key adx_CloudFlowConsumer_adx_webrole kept as-is
• statecode/statuscode defaults omitted; empty strings written as ''
• Fields alphabetically sorted

6.5 Git Commit

After all YAML files (and any new web roles) are created, do a single git commit. Include the HTML plan file (use the actual output path from the render script's JSON response).

Output: One .cloudflowconsumer.yml per new flow, committed (skipped for integration-only flows)

Phase 7: Client-Side Integration

Goal: Generate typed service code that calls each registered flow from the site's frontend

Cloud flows are invoked via an authenticated POST to /_api/cloudflow/v1.0/trigger/<workflowEntityId>. The request must include a CSRF token fetched from /_layout/tokenhtml, and the payload must be wrapped in an eventData key as a JSON-stringified string.

Actions:

7.1 Discover Existing Frontend Patterns

Use the Explore agent to scan the site:

"Check the Power Pages site frontend for existing patterns. Look for:

1. Any existing API service files in src/shared/, src/services/, src/api/, or similar
2. Existing CSRF token handling (__RequestVerificationToken, _layout/tokenhtml)
3. The framework (React / Vue / Angular / Astro) and its conventions (hooks, composables, services)
4. Any existing cloud flow calls (/_api/cloudflow/) already in the codebase — note which flows are already called and from which pages/components Report all findings."

For integration-only flows, the Explore agent's findings in point 4 are critical — they reveal where the flow is already wired in, so the new integration targets a different page/component without duplicating existing call sites.

7.2 Create or Update the Cloud Flow Service

Do not use the site's existing OData fetch wrapper for cloud flow triggers.

If the site has a centralized API client (e.g. powerPagesFetch) that targets the Dataverse Web API, do not use it here. Cloud flow trigger URLs (/_api/cloudflow/v1.0/trigger/...) are a different endpoint that does not accept OData-specific request headers. Using an OData wrapper will cause the request to fail.

Use direct fetch or an equivalent low-level HTTP client (for example, Angular's HttpClient) for cloud flow calls. Do not use any OData-specific wrappers or helpers that automatically add Dataverse Web API headers. You may reuse the site's existing CSRF token helper if one is exported — but perform the HTTP request via fetch or the chosen low-level client directly.

Based on the framework and existing patterns:

If the site already has an API service layer: add the new flow functions to it. If no service layer exists: create src/services/cloudFlowService.{ts,js} (or framework equivalent).

For integration-only flows: if a trigger function already exists in the service file, do not create a duplicate — reuse it. Only add a new trigger function if the flow has no existing function in the codebase.

Service structure

REQUIRED for every flow trigger call — without these headers the request returns a 500 error:

• __RequestVerificationToken: CSRF token fetched from /_layout/tokenhtml
• X-Requested-With: XMLHttpRequest

The payload must be wrapped in eventData as a JSON-stringified string — this is the shape the Power Pages cloud flow endpoint expects:

Copy & paste — that's itCopy

{ "eventData": "{\"Email\":\"abc@contoso.com\"}" }

The service file must contain a getCsrfToken helper and one trigger function per flow. If a getCsrfToken helper already exists in the codebase, reuse it; otherwise create it. Either way, every trigger function must call it and include both headers shown above.

CSRF helper (reuse existing or create):

async function getCsrfToken(): Promise<string> {
  const res = await fetch('/_layout/tokenhtml');
  const html = await res.text();
  const match = html.match(/value="([^"]+)"/);
  if (!match) throw new Error('CSRF token not found');
  return match[1];
}

Per-flow trigger function (one per registered flow):

// Trigger: <flow display name>
// API: /_api/cloudflow/v1.0/trigger/<workflowEntityId>
export async function <camelCaseFunctionName>(payload: Record<string, unknown> = {}): Promise<unknown> {
  const token = await getCsrfToken();
  const response = await fetch('/_api/cloudflow/v1.0/trigger/<workflowEntityId>', {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      '__RequestVerificationToken': token,
      'X-Requested-With': 'XMLHttpRequest',
    },
    body: JSON.stringify({ eventData: JSON.stringify(payload) }),
  });
  if (!response.ok) {
    throw new Error(`Cloud flow trigger failed: ${response.status} ${response.statusText}`);
  }
  // 202 Accepted = flow has no Response action (fire-and-forget); 200 OK = flow returns data
  if (response.status === 202) return null;
  return response.json();
}

Framework-specific patterns:

Group all flow functions together in one file. Do not create a separate file per flow.

7.3 Wire into UI

Read the relevant pages/components, add call sites, loading states, success/error feedback. Replace any placeholder or mock handlers that are meant to be backed by the flows.

For integration-only flows, this is the primary deliverable of the entire skill run. Identify the new target page or component where the flow should be integrated (based on the user's request and the scenario from Phase 3.3), import the existing trigger function, and wire it in with proper loading/error handling. Do not modify existing call sites that already work.

7.4 Git Commit

Commit service file and any UI changes.

Output: cloudFlowService.{ts,js} (or equivalent) created/updated, UI wired

Phase 8: Verify & Summarize

Goal: Validate all created YAMLs, record skill usage, present summary, offer deployment

Actions:

8.1 Validate Each YAML

Skip for integration-only flows — their YAML was not created or modified by this run. Only validate YAMLs for new flows.

For each created .cloudflowconsumer.yml, verify:

• id is a valid UUID
• processid matches the flow's workflowEntityId
• adx_CloudFlowConsumer_adx_webrole is non-empty
• All role UUIDs in the array correspond to existing .webrole.yml files
• name equals the flow's display name
• No adx_-prefixed scalar fields (no adx_processid, adx_websiteid, adx_name)
• statecode and statuscode are absent

8.1b Validate Client-Side Integration

Read the service file created in Phase 7 and verify:

• Service file exists at the expected path (e.g. src/services/cloudFlowService.ts or framework equivalent)
• A getCsrfToken helper is present — either defined in the file or imported from another module
• Every registered flow has a corresponding exported trigger function
• Each trigger function calls getCsrfToken() and sets both required headers:
  • __RequestVerificationToken set to the CSRF token value
  • X-Requested-With: XMLHttpRequest
• Each trigger function's fetch URL exactly matches /_api/cloudflow/v1.0/trigger/<workflowEntityId> for the correct flow
• The request body wraps the payload as JSON.stringify({ eventData: JSON.stringify(payload) }) — not a bare JSON.stringify(payload)
• No OData-specific wrapper (e.g. powerPagesFetch) is used — raw fetch only
• At least one UI component or page imports and calls the trigger function(s)

If any check fails, fix the issue in the service file or UI before continuing.

8.2 Record Skill Usage

Reference: ${PLUGIN_ROOT}/references/skill-tracking-reference.md

Use --skillName "AddCloudFlow".

8.3 Present Summary

8.4 Ask to Deploy

🚦 Gate (plan · add-cloud-flow:8.4.deploy): Post-implementation deploy prompt — flows don't trigger until deployed.

Trigger: All flow YAML + client integration in place. Why we ask: Auto-deploy picks wrong env. Cancel leaves: Nothing — artifacts stay on disk; no deploy fired.

Use AskUserQuestion:

🚦 Gate (plan · add-cloud-flow:8.4.test): Post-deploy validation prompt — invokes /test-site to confirm the flow trigger endpoint returns 202/200.

Trigger: Deploy from the previous gate succeeded. Why we ask: Skipping is harmless (manual test still possible); auto-invoking /test-site adds runtime and Playwright traffic. Cancel leaves: Nothing — deploy has already completed.

If "Yes": Invoke /deploy-site. After it succeeds, use AskUserQuestion:

If "Yes, run tests": Invoke /test-site with context: test the scenario where the flow API is triggered — perform the action on the site (e.g. submit the form, click the button) that calls /_api/cloudflow/v1.0/trigger/<workflowEntityId>, confirm the request returns 202 Accepted or 200 OK, and check Power Automate run history to confirm the flow executed.

If "No": Remind the user that flows and client-side calls won't work until deployed, and suggest running /test-site later to validate the flow trigger works correctly in the deployed environment.

8.5 Post-Deploy Notes

• Flow trigger URL: Populated automatically by the portal runtime after deploy. Check Power Pages design studio to confirm.
• Testing: Use /test-site to verify the integration. Sign in as a user with the assigned role, trigger the flow action, confirm /_api/cloudflow/v1.0/trigger/<id> returns 202/200, and check Power Automate run history.
• Anonymous flows: Test in an incognito window.
• Role changes: Edit adx_CloudFlowConsumer_adx_webrole in the YAML, then redeploy.
• Adding more flows later: Run this skill again — new flows are available for full setup, and already-registered flows are available for additional frontend integration on other pages.

Important Notes

Throughout All Phases

• Multiple flows: Process all flows in the approved plan before committing or moving to Phase 7
• Minimum roles: Never assign all roles by default; use only what the scenario requires
• Anonymous Users: Always surface explicitly in plan and confirm step
• Trigger URL blank at creation: Expected — set by the portal runtime at deploy time
• fileSlug vs flowName: --fileSlug is the filename only; --flowName is the name YAML field
• Client service file: Group all flows into one service file, not one file per flow; always wire into the UI without asking
• CSRF headers are mandatory: Every flow trigger function must call getCsrfToken() and send both __RequestVerificationToken and X-Requested-With: XMLHttpRequest — omitting either causes a 500 error
• Payload shape: The request body must be JSON.stringify({ eventData: JSON.stringify(payload) }) — the cloud flow endpoint expects the payload nested under an eventData key as a double-stringified JSON string
• Integration-only flows: Already-registered flows skip Phases 4–6 (roles and metadata). Their primary deliverable is new frontend call sites in Phase 7. Do not modify existing .cloudflowconsumer.yml files or web role assignments for these flows. Reuse existing trigger functions when they already exist in the service file — only add new UI call sites.
• No flows available: When no flows with a PowerPages trigger exist in the environment (zero results from API), stop the workflow and direct the user to create a flow in Power Automate with the "When a Power Pages flow step is run" trigger first.

Key Decision Points (Wait for User)

1. Phase 1.3: Deploy now or cancel (if .powerpages-site missing)
2. Phase 3.1: Which flows to add or integrate (new and/or already-registered)
3. Phase 3.3: Clarify scenario / audience per flow (if unclear)
4. Phase 5.3: Approve HTML plan before writing any files
5. Phase 8.4: Deploy now or later

Progress Tracking

Before starting Phase 1, create a task list with all phases using TaskCreate:

Mark each task in_progress when starting and completed when done via TaskUpdate.

Begin with Phase 1: Verify Prerequisites