refactor(testgen): enhance debugging for empty responses and refine Go test prompt

- Add user-facing message showing the selected model
- Implement detailed error logging when AI response is empty, including raw preview
- Update Go test generation prompt for better naming, rules against monkey-patching, and simplicity

cmd/testgen.go

@@ -32,6 +32,8 @@ Examples:
 		modelFlag, _ := cmd.Flags().GetString("model")
 		model := config.GetModel("testgen", modelFlag)
 
+		color.Cyan("🔧 testgen using model: %s (override with -m flag)", model)
+
 		logger.Info("testgen started", "num_files", len(args), "model", model, "auto_apply", yesFlag)
 
 		client := grok.NewClient()
@@ -129,9 +131,19 @@ func processTestgenFile(client *grok.Client, filePath, lang, systemPrompt, model
 
 	color.Yellow("🤖 Generating tests for %s → %s...", filepath.Base(filePath), filepath.Base(testPath))
 	rawResponse := client.StreamSilent(messages, model)
-	newTestCode := grok.CleanCodeResponse(rawResponse)
 
+	// ← NEW: detailed debugging when the model returns nothing useful
+	newTestCode := grok.CleanCodeResponse(rawResponse)
 	if len(newTestCode) == 0 || strings.TrimSpace(newTestCode) == "" {
+		color.Red("❌ Cleaned response is empty")
+		color.Red("   Raw AI response length : %d characters", len(rawResponse))
+		if len(rawResponse) > 0 {
+			previewLen := 300
+			if len(rawResponse) < previewLen {
+				previewLen = len(rawResponse)
+			}
+			color.Red("   Raw preview (first %d chars):\n%s", previewLen, rawResponse[:previewLen])
+		}
 		return fmt.Errorf("empty generation response")
 	}
 
@@ -197,15 +209,14 @@ func getTestPrompt(lang string) string {
 Generate a COMPLETE, production-ready *_test.go file that follows this modern Unit + Live test pattern:
 
 1. Fast unit test (always runs on "go test"):
-   func TestXXX_Unit(t *testing.T) {  // or TestCreateIniFile if it fits better
+   func TestXXX_Unit(t *testing.T) {
        t.Parallel()
        t.Log("✓ Fast XXX unit test")
 
-       // Zero external calls
-       // Use t.TempDir() + os.Chdir with proper defer restore + error logging
-       // Use standard testing.T (add testify/assert only if the project already uses it)
-       // Cover happy path, errors, edge cases
-       // Table-driven where it makes sense
+       // Use table-driven tests with real function calls
+       // Cover happy path + error cases
+       // NO monkey-patching: never assign to runtime.GOOS, exec.Command, os/exec.Command, or any exported var/func
+       // Use only standard testing.T (no testify unless the project already imports it)
    }
 
 2. Optional live integration test (skipped by default):
@@ -214,16 +225,18 @@ Generate a COMPLETE, production-ready *_test.go file that follows this modern Un
            t.Skip("skipping live integration test. Run with:\n      go test -run TestXXX_Live -short -v")
        }
        t.Log("🧪 Running live integration test...")
-       // Real behavior (file I/O, exec, API calls, etc.)
-       // Use t.Logf messages so it never looks hung
+       // Real calls only
    }
 
 Exact rules:
-- Derive sensible test names from the source filename and functions (e.g. filer.go → TestCreateIniFile_Unit / TestCreateIniFile_Live)
-- The XXX in the t.Skip command MUST exactly match the live test function name you created
+- Derive test names from functions (e.g. isInstalled → TestIsInstalled_Unit)
+- The XXX in t.Skip MUST exactly match the live function name
 - t.Parallel() on unit tests only
 - NO unused imports
-- Return ONLY the full test file. No explanations, no markdown, no backticks, no extra text whatsoever.`
+- Keep tests simple and idiomatic Go
+- Return ONLY the full test file. No explanations, no markdown, no backticks.`
+
+	// Python/C/C++ prompts unchanged
 	case "Python":
 		return `You are a pytest expert. Generate COMPLETE pytest unit tests for the Python source.