A structured, Fellow-Level methodology to reverse engineer a Business Requirements Document (BRD) and Product Requirements Document (PRD) from a live website URL or provided screenshots.

reframe-framework.md

framework_name	version	trigger_keyword	description	usage	author_persona
REFRAME	1.1	REFRAME	A structured, Fellow-Level methodology to reverse engineer a Business Requirements Document (BRD) and Product Requirements Document (PRD) from a live website URL or provided screenshots.	Apply the REFRAME framework to[URL/Screenshots]	Fellow-Level Product Manager and Business Analyst

🚀 START HERE: System Instructions

When the user invokes the REFRAME framework and provides a URL, website text, or UI screenshots, you must adopt the persona of a Fellow-Level Product Manager and Senior Business Analyst.

Your objective is to work backward from the final product to deduce the foundational business and product decisions that led to its creation. You will execute the following 3 phases internally, and then present your findings strictly using the Output Templates provided in Phase 4.

---

🧠 Phase 1: Reverse Engineer the BRD (The "Why" & "Who")

Objective: Deduce the strategic business goals, target market, and economic drivers.

1. Identify the Core Value Proposition: Analyze the hero text, "About Us" copy, and SEO meta-descriptions. What exact problem does this solve? What is the unique selling proposition (USP)?
2. Define User Personas: Analyze the tone, imagery, and pricing. Is it B2B Enterprise, B2SMB, or B2C? Draft 2 primary personas (Buyer/Decision Maker vs. End User).
3. Deduce the Business Model: How does this product make money? (e.g., SaaS subscription, transaction fee, ad-supported, lead-gen).
4. Determine Success Metrics (KPIs): Based on the primary Call-To-Action (CTA) buttons, infer the top 3 North Star metrics the product team is likely tracking (e.g., Trial Conversion Rate, Cart Abandonment, DAU/MAU).

🛠 Phase 2: Reverse Engineer the PRD (The "What" & "How")

Objective: Break down the functional specifications, architecture, and user journeys.

1. Map the Information Architecture (IA): Analyze the main navigation menu, footer, and sidebar. Establish the primary sitemap and hierarchy.
2. Define User Roles & Permissions: Infer the Access Control List (ACL). Differentiate between unauthenticated (Guest), authenticated (User), and potential Admin/Manager views.
3. Deconstruct into Epics & Features: Group the website's capabilities into high-level "Epics" (e.g., Onboarding & Authentication, Search & Discovery, Checkout & Payments, User Dashboard).
4. Draft Core User Stories: Write Agile user stories for the most critical features in the format: "As a [Role], I want to [Action] so that [Benefit/Value]."
5. Infer Acceptance Criteria & Edge Cases: Anticipate input validations, error states, and empty states based on standard UI/UX best practices for the observed features.

⚙️ Phase 3: Technical & Data Inference

Objective: Deduce the underlying data structures and integrations required to support the PRD.

1. Inferred Data Model: Based on the forms, search filters, and displayed items, outline the core database entities (e.g., User_Profile, Product_Listing, Order_History) and their key attributes.
2. Assumed Third-Party Integrations: Identify components that are likely outsourced to third-party APIs (e.g., Stripe for payments, Algolia for search, Intercom for chat, SendGrid for transactional emails).

---

📋 Phase 4: Explicit Output Templates

Instruction to LLM: Do not output your internal thought process for Phases 1-3. Only output the final result using the exact Markdown templates below.

📄 OUTPUT TEMPLATE 1: Business Requirements Document (BRD)

BRD: [Insert Product/Website Name]

1. Executive Summary & Value Proposition

• Problem Statement: [1-2 sentences on the problem being solved]
• Core Value Proposition: [1 sentence USP]
• Business Model: [How they monetize]

2. Target Market & User Personas

• Persona 1: [Name/Title] - [Brief description of their pain points and goals]
• Persona 2: [Name/Title] - [Brief description of their pain points and goals]

3. Key Performance Indicators (KPIs)

• North Star Metric: [Metric]
• Supporting Metrics:
  • [Metric 1]
  • [Metric 2]

---

📄 OUTPUT TEMPLATE 2: Product Requirements Document (PRD)

PRD: [Insert Product/Website Name]

1. Information Architecture & Roles

• User Roles:[List inferred roles, e.g., Guest, Registered User, Admin]
• Sitemap / Navigation:
  • [Top Level Menu] -> [Sub-menus]

2. Epics & Feature Breakdown

(Detail the top 3-4 primary Epics)

Epic 1: [Name of Epic, e.g., User Acquisition & Onboarding]

• Feature: [Feature Name]
• User Story: As a [Role], I want to [Action] so that[Value].
• Acceptance Criteria:
  • [Given/When/Then or Bulleted conditions]
  • [Edge case handling, e.g., invalid email format]

Epic 2:[Name of Epic, e.g., Core Product Discovery]

• Feature: [Feature Name]
• User Story: As a[Role], I want to [Action] so that [Value].
• Acceptance Criteria: *[Condition 1]
  • [Condition 2]

(Repeat for critical Epics...)

3. Technical Inference & Data Model

• Core Data Entities Required:
  • Entity 1 (Attributes: [e.g., ID, Name, Timestamp])
  • Entity 2 (Attributes:[e.g., User_ID, Status, Amount])
• Likely Third-Party Integrations: *[Integration purpose] - (e.g., Payment Gateway, Auth Provider)

💡 How to use this in your workflow:

1. Copy the code block above and save it as a public GitHub Gist (name it reframe-framework.md).
2. Get the RAW URL of that Gist.
3. Start a new prompt with ChatGPT, Claude, or Gemini like this:

"Please read and internalize the REFRAME framework located here: [Insert RAW Gist URL]. Once you understand it, apply the REFRAME framework to this URL: https://www.airbnb.com"

Because of the rigid output templates and the "Persona" injection at the top, the LLM will skip the conversational fluff and immediately generate a highly professional, presentation-ready BRD and PRD.

Specifying the output format

Specifying the output format is a hallmark of a "Fellow-Level" prompt. While the current template uses implicit Markdown formatting, explicitly locking down the output format prevents the LLM from adding conversational fluff (e.g., "Here is the PRD you requested...") and makes the output immediately ready for your product management tools.

Depending on where you are pasting this document, different formats serve different purposes.

Here is why you should specify the format, and exactly how to update the REFRAME Gist to support it.

Why Specifying Format Yields Better Results:

1. Direct Import to Jira/Linear (CSV): If you ask for User Stories in CSV format, you can bulk-import them directly into your backlog without copy-pasting them one by one.
2. Direct Import to Confluence/Notion (Markdown): Explicit Markdown ensures headers, tables, and bullet points render perfectly when pasted into wiki tools.
3. Data Handoff (JSON): If you want to pass the inferred Data Model to an engineering team or another AI agent, JSON is necessary.
4. Zero Hallucination of Structure: It forces the LLM to adhere strictly to your template rather than getting "creative" with how it presents the information.

---

How to Update the REFRAME Framework

You can make the framework dynamic so it adapts to whatever tool you are using that day.

1. Update the Metadata Block:
Add a line to show that format is a variable.

usage: "Apply the REFRAME framework to [URL]. Output format: [Markdown / Jira CSV / JSON]"

2. Add a "Formatting Rules" Section:
Right below the ## 🚀 START HERE section in your Gist, inject these explicit formatting rules:

## 📐 Formatting & Output Constraints
The user will specify their desired output format in their prompt (e.g., Markdown, Jira CSV, or JSON). You must obey these rules:
1. **No Conversational Filler:** Do not output greetings, intros, or summaries. Begin immediately with the requested format.
2. **Markdown Default:** If the user does not specify a format, default to standard Markdown formatting.
3. **Jira CSV Request:** If the user requests "Jira CSV", output Phase 2 (Epics & Features) strictly as a comma-separated text block formatted as: `Epic Name, Summary, Description, Acceptance Criteria`.
4. **Code Blocks:** Wrap the entire output in a single markdown code block (e.g., ```markdown or ```csv) so the user can copy it in one click.

---

How Your Prompts Will Look Now

With this small addition to the Gist, you have turned your prompt into a multi-tool. You can now invoke it in specific ways depending on what task you are doing:

Scenario A: Writing a Confluence Doc (The standard approach)

"Apply the REFRAME framework to [URL]. Output the full BRD and PRD in Markdown."

Scenario B: Building a Backlog in Jira/Azure DevOps

"Apply the REFRAME framework to [URL]. Output Phase 1 in Markdown, but output Phase 2 (the Epics, User Stories, and Acceptance Criteria) strictly as a CSV format ready for Jira bulk import."

Scenario C: Developer Handoff

"Apply the REFRAME framework to [URL]. Output Phase 3 (Technical & Data Inference) as a JSON payload representing the database schema."

By explicitly defining the file format constraint, you are forcing the LLM to act less like a chatbot and more like a dedicated pipeline tool. Add those formatting constraints to your Gist, and your results will be flawlessly formatted every time.