feat(mcp): add guide_testing endpoint with TestKit bindings reflection

src/TALXIS.CLI.MCP/GuideReasoningEngine.cs

@@ -23,6 +23,7 @@ public class GuideReasoningEngine
         ["guide_deployment"] = ["deployment-sequence", "solution-management"],
         ["guide_data"] = ["data-migration-workflow"],
         ["guide_config"] = [],
+        ["guide_testing"] = ["testing-workflow"],
     };
 
     /// <summary>

src/TALXIS.CLI.MCP/Program.cs

@@ -30,6 +30,10 @@
 var publicSkillLoader = new PublicSkillLoader();
 publicSkillLoader.LoadIndex();
 
+// Load UI testing step bindings catalog via reflection
+var testingBindingsCatalog = new TestingBindingsCatalog();
+testingBindingsCatalog.Load();
+
 // Session-scoped active tool set — starts with always-on tools only
 var activeToolSet = new ActiveToolSet();
 var guideHandler = new GuideHandler(mcpToolRegistry.Catalog, activeToolSet, reasoningEngine);
@@ -71,6 +75,7 @@
 - guide_deployment: Deployment lifecycle — import/export/pack solutions, manage components, publish. Requires profile.
 - guide_data: LIVE data operations — SQL/FetchXML/OData queries, record CRUD, bulk ops, CMT migration. Requires profile.
 - guide_config: CLI setup — auth credentials, connections, profiles, settings. Required before environment operations.
+- guide_testing: UI test generation — discover available Reqnroll step bindings for Power Apps BDD tests.
 
 WORKFLOW: Call a guide tool → use execute_operation for immediate execution → discovered tools become direct calls on next turn.
 
@@ -214,6 +219,10 @@ async ValueTask<CallToolResult> HandleGuideToolAsync(
             };
         }
     }
+    else if (guideName == "guide_testing")
+    {
+        result = await HandleGuideTestingAsync(query, top, server, ct);
+    }
     else if (workflowScope is not null)
     {
         result = await guideHandler.HandleWorkflowGuideAsync(workflowScope, query, top, server, ct, guideName);
@@ -228,6 +237,78 @@ async ValueTask<CallToolResult> HandleGuideToolAsync(
     return result;
 }
 
+// Handles guide_testing calls — uses TestingBindingsCatalog + sampling to recommend step bindings
+async ValueTask<CallToolResult> HandleGuideTestingAsync(
+    string query, int top, McpServer server, CancellationToken ct)
+{
+    if (testingBindingsCatalog.Count == 0)
+    {
+        return new CallToolResult
+        {
+            Content = [new TextContentBlock { Text = "No step bindings loaded. Ensure the TALXIS.TestKit.Bindings assembly is available." }],
+            IsError = true
+        };
+    }
+
+    if (string.IsNullOrEmpty(query))
+    {
+        // No query — return full catalog listing
+        return new CallToolResult
+        {
+            Content = [new TextContentBlock { Text = testingBindingsCatalog.GetCatalogPrompt() }]
+        };
+    }
+
+    // Use sampling to select relevant bindings based on user's query
+    var skillsContext = reasoningEngine.GetSkillsContext("guide_testing");
+    var catalogPrompt = testingBindingsCatalog.GetCatalogPrompt();
+
+    var systemPrompt = $@"You are a Power Apps UI test automation assistant. Given the user's testing task and available step bindings, produce a Gherkin feature file or scenario.
+
+FORMAT YOUR RESPONSE AS:
+1. A complete Gherkin scenario (or scenarios) using the available step bindings
+2. Include comments explaining any custom steps that would need to be implemented
+
+RULES:
+- Use ONLY the step bindings listed below when possible
+- For login, ALWAYS start with: Given I am logged in to the '{{app}}' app as '{{user}}'
+- Use realistic placeholder values based on the user's description
+- Each scenario should test ONE behavior
+- Include test data setup (Given steps) before actions (When steps)
+- End with assertions (Then steps)
+- If the available bindings don't cover something, note it as a custom step with a comment
+
+{catalogPrompt}{skillsContext}";
+
+    var samplingParams = new CreateMessageRequestParams
+    {
+        Messages =
+        [
+            new SamplingMessage
+            {
+                Role = Role.User,
+                Content = [new TextContentBlock { Text = $"Generate a Gherkin test for this scenario: {query}" }]
+            }
+        ],
+        SystemPrompt = systemPrompt,
+        MaxTokens = 2000,
+        ModelPreferences = new ModelPreferences
+        {
+            SpeedPriority = 0.6f,
+            CostPriority = 0.4f,
+            IntelligencePriority = 0.8f
+        },
+    };
+
+    var result = await server.SampleAsync(samplingParams, ct);
+    var responseText = result.Content.OfType<TextContentBlock>().FirstOrDefault()?.Text ?? "";
+
+    return new CallToolResult
+    {
+        Content = [new TextContentBlock { Text = responseText }]
+    };
+}
+
 // Bridge: execute_operation dispatches any tool from the internal catalog
 async ValueTask<CallToolResult> HandleExecuteOperationAsync(
     IDictionary<string, JsonElement>? arguments, RequestContext<CallToolRequestParams> ctx, CancellationToken ct)
@@ -473,7 +554,7 @@ async ValueTask<CallToolResult> ExecuteAsTaskAsync(
 bool IsGuideTool(string toolName)
 {
     return toolName is "guide" or "guide_workspace" or "guide_environment"
-        or "guide_deployment" or "guide_data" or "guide_config";
+        or "guide_deployment" or "guide_data" or "guide_config" or "guide_testing";
 }
 
 // Helper: checks if a tool is an MCP-specific in-process tool (not a CLI subprocess)
@@ -563,6 +644,13 @@ void RegisterAlwaysOnTools(ActiveToolSet toolSet, McpToolRegistry registry, Publ
         InputSchema = BuildGuideInputSchema()
     });
 
+    toolSet.AddAlwaysOn(new Tool
+    {
+        Name = "guide_testing",
+        Description = @"Helps generate Power Apps UI tests using Reqnroll (BDD). Discovers available step bindings from TALXIS.TestKit.Bindings and generates Gherkin scenarios. Provide a description of what you want to test and get ready-to-use feature file content with Given/When/Then steps.",
+        InputSchema = BuildGuideInputSchema()
+    });
+
     // execute_operation bridge — for same-turn execution
     toolSet.AddAlwaysOn(new Tool
     {

src/TALXIS.CLI.MCP/Skills/Internal/testing-workflow.md

@@ -0,0 +1,69 @@
+# UI Testing Workflow
+
+<!-- Internal reasoning skill: contains ONLY test-generation guidance. -->
+<!-- For available step bindings, see the TestingBindingsCatalog prompt. -->
+
+## User wants to "write a UI test" / "test a form" / "test navigation"
+
+-> STRUCTURE: Feature file (.feature) with Scenario(s) using Given/When/Then
+-> ALWAYS start with a Given step for login: `Given I am logged in to the '{app}' app as '{user}'`
+-> ALWAYS use pre-built step bindings from TALXIS.TestKit.Bindings where available
+-> ONLY write custom step bindings for app-specific logic not covered by the library
+
+## Test Structure Best Practices
+
+-> Feature files group related scenarios by business capability
+-> Each scenario should be independent (no shared state between scenarios)
+-> Use Background for shared Given steps across all scenarios in a feature
+-> Keep scenarios focused on ONE behavior/assertion
+-> Use Scenario Outline for data-driven tests
+
+## Given/When/Then Conventions
+
+-> Given: Setup preconditions (login, create test data, navigate to starting point)
+-> When: Perform the action being tested (click, enter data, navigate)
+-> Then: Assert the expected outcome (field values, visibility, error messages)
+-> Avoid multiple When steps — split into separate scenarios instead
+
+## Test Data Setup
+
+-> Use `Given I have created '{alias}'` with JSON data files in a /data folder
+-> Data files use Web API deep-insert syntax with @logicalName, @alias, @extends
+-> Use faker.js templates for dynamic data ({{name.firstName}}, {{finance.amount}})
+-> Set `deleteTestData: true` in appsettings.json for cleanup after scenarios
+
+## Common Patterns
+
+### Testing form field entry:
+```gherkin
+When I enter '{value}' into the '{field label}' field on the form
+```
+
+### Testing navigation:
+```gherkin
+When I open the sub area '{subarea}' under the '{area}' area
+When I open the '{subarea}' sub area of the '{group}' group
+```
+
+### Testing command bar:
+```gherkin
+When I select the '{command}' command
+Then I should be able to see the '{command}' command
+```
+
+### Testing grids/views:
+```gherkin
+When I open the record at position '{n}' in the grid
+Then I can see '{n}' records in the grid
+```
+
+### Testing lookups:
+```gherkin
+When I select '{value}' from the '{field}' lookup field
+```
+
+## Error Recovery
+
+-> If a step binding fails with a timeout: check if Driver.WaitForTransaction() is needed
+-> If login fails: verify user credentials in appsettings.json and OTP token configuration
+-> If element not found: the app may need WaitForPageToLoad before interaction

src/TALXIS.CLI.MCP/TALXIS.CLI.MCP.csproj

@@ -47,6 +47,7 @@
   <ItemGroup>
     <PackageReference Include="Microsoft.Extensions.Hosting" Version="10.0.5" />
     <PackageReference Include="ModelContextProtocol" Version="1.2.0" />
+    <PackageReference Include="TALXIS.TestKit.Bindings" Version="1.0.10" ExcludeAssets="build;buildTransitive" />
     <ProjectReference Include="../TALXIS.CLI/TALXIS.CLI.csproj" />
     <ProjectReference Include="../TALXIS.CLI.Features.Data/TALXIS.CLI.Features.Data.csproj" />
     <ProjectReference Include="../TALXIS.CLI.Features.Docs/TALXIS.CLI.Features.Docs.csproj" />