refactor(recipe): enhance parsing for multi-line step content

Replace regex-based sub-section extraction with label-indexed parsing to
handle multi-line content, blank lines, and ensure reliability without
duplicates. Add clarifying comments for templates and summary extraction.

internal/recipe/loader.go

@@ -12,8 +12,8 @@ import (
 )
 
 var (
+	// stepRe still finds the headings (this one is solid)
 	stepRe = regexp.MustCompile(`(?m)^### Step (\d+): (.+)$`)
-	subRe  = regexp.MustCompile(`(?m)^(\*\*(?:Objective|Instructions|Expected output):\*\*)\s*(.+?)(?:\n\n|\n###|\z)`)
 )
 
 func Load(path string, userParams map[string]any) (*Recipe, error) {
@@ -45,7 +45,7 @@ func Load(path string, userParams map[string]any) (*Recipe, error) {
 		}
 	}
 
-	// Render templates
+	// Render templates (so {{.package_path}} becomes "internal")
 	tpl, err := template.New("recipe").Parse(string(parts[2]))
 	if err != nil {
 		return nil, err
@@ -61,7 +61,7 @@ func Load(path string, userParams map[string]any) (*Recipe, error) {
 		r.Overview = strings.TrimSpace(body[:idx])
 	}
 
-	// Extract steps — split-based to guarantee no duplicates
+	// Extract steps with robust multi-line parsing
 	matches := stepRe.FindAllStringSubmatch(body, -1)
 	for i, m := range matches {
 		stepNum := i + 1
@@ -77,20 +77,42 @@ func Load(path string, userParams map[string]any) (*Recipe, error) {
 		section := body[start:end]
 
 		step := Step{Number: stepNum, Title: title}
-		for _, sub := range subRe.FindAllStringSubmatch(section, -1) {
+
-			switch sub[1] {
+		// Simple, reliable label-based parsing (handles multi-line + blank lines)
+		labels := []string{"**Objective:**", "**Instructions:**", "**Expected output:**"}
+		for _, label := range labels {
+			labelStart := strings.Index(section, label)
+			if labelStart == -1 {
+				continue
+			}
+			contentStart := labelStart + len(label)
+			contentEnd := len(section)
+
+			// Find next label or end of section
+			for _, nextLabel := range labels {
+				next := strings.Index(section[contentStart:], nextLabel)
+				if next != -1 {
+					contentEnd = contentStart + next
+					break
+				}
+			}
+
+			content := strings.TrimSpace(section[contentStart:contentEnd])
+
+			switch label {
 			case "**Objective:**":
-				step.Objective = strings.TrimSpace(sub[2])
+				step.Objective = content
 			case "**Instructions:**":
-				step.Instructions = strings.TrimSpace(sub[2])
+				step.Instructions = content
 			case "**Expected output:**":
-				step.Expected = strings.TrimSpace(sub[2])
+				step.Expected = content
 			}
 		}
+
 		r.Steps = append(r.Steps, step)
 	}
 
-	// Final summary = everything after the last step
+	// Final summary (everything after last step)
 	if len(matches) > 0 {
 		lastMatch := matches[len(matches)-1][0]
 		lastIdx := strings.LastIndex(body, lastMatch)