Skip to main content
Writing a product spec manually takes hours and still ends up inconsistent. Many teams skip or rush it. And when specs are weak, the wrong product gets built, the design lacks context, and PMs spend more time in meetings than shipping. This prompt fixes that. Just fill in 4 inputs about your product and get a structured, team-ready spec in seconds, complete with personas, user stories, requirements, a timeline, and a risk assessment, every time.

How the prompt works

The system prompt gives the LLM a clear role: that of an expert Product Manager. It then requires six specific sections in every response. This structure is intentional. Without it, models default to a generic summary rather than a document a team can act on. The user prompt is kept short; just four variables that capture what any product needs without a long brief.

System prompt

Sets the LLM’s role and defines the six sections it must always include in the output.

User prompt

The four variables a user fills in to describe their product. Kept short so it’s easy to use without a long brief.
Product Specification: FoodSnap - AI Recipe GeneratorExecutive SummaryFoodSnap is an AI-powered mobile app that transforms ingredient photos into personalized recipe recommendations, targeting busy professionals who want quick, tailored cooking solutions.Target User AnalysisPrimary Persona: Busy Professional Cook (Age 25-40)
  • Goals: Quick meal ideas, use existing ingredients, healthy eating
  • Pain Points: Lack of cooking inspiration, food waste, time constraints
Feature Specifications
  1. AI Ingredient Recognition
    • User Story: “As a user, I want to snap a photo of my fridge contents and get recipe suggestions”
    • Requirements: Computer vision API, ingredient database, accuracy over 90%
  2. Personalized Recipe Engine
    • User Story: “As a user, I want recipes that match my dietary preferences and skill level”
    • Requirements: User preference profiling, recipe difficulty scoring, dietary filters
  3. Smart Shopping Lists
    • User Story: “As a user, I want to generate shopping lists for missing ingredients”
    • Requirements: Recipe parsing, inventory tracking, grocery store integration
Technical Requirements
  • Platform: iOS/Android native apps
  • Backend: Cloud-based ML models, recipe database API
  • Performance: Under 3 second photo processing, offline recipe viewing
Success Metrics
  • User Retention: 80% after 30 days
  • Engagement: 3+ recipes generated per user weekly
  • Satisfaction: 4.5+ app store rating
  • Conversion: 25% of users upgrade to premium within 60 days
Implementation Timeline
  • Phase 1 (Months 1-2): Core photo recognition, basic recipe database
  • Phase 2 (Months 3-4): Personalization engine, dietary filters
  • Phase 3 (Months 5-6): Shopping list integration, premium features
Risk Assessment
  • Technical: AI accuracy may vary with photo quality | Mitigation: Extensive training data, user feedback loop
  • Market: Competitive recipe app landscape | Mitigation: Focus on unique photo-to-recipe workflow

Import into Adaline

This prompt comes with a ready-to-import Adaline project file. It includes the prompt, dataset, and evaluators, all pre-configured.

Evaluations and dataset

Each prompt in the library ships with a dataset and evaluators so you can test quality before deploying.

Evaluators

This use case comes with three evaluators: two LLM-as-a-judge prompts and one performance rule. Copy any to use in your own eval pipeline.

Structural completeness

Checks that all six required sections are present in every output.

Actionability and specificity

Checks that the output is specific enough for a team to act on immediately.

Response length

Guards against bloated output. A good spec is thorough but still easy to scan.

Dataset

Four product types across different industries, each row maps directly to the four variables in the user prompt.