AIStoryMapper

This project was created to support Technical Program Managers (TPMs), Product Owners, and Engineers in accelerating the gap between customer feedback and actionable engineering work.

Table of Contents

• Overview
• Prerequisites
• Installation
• Features
• Use Cases
• Usage
• Future Improvements

Overview

AIStoryMapper is a developer productivity tool that extracts structured Agile work items — including user stories and spikes — directly from meeting transcripts or project notes.

Powered by GitHub Copilot Chat (GHC), this tool supports two usage modes:

• Custom Agent Mode (for VS Code Insiders with agent support)
• Prompt-based Mode (for all other users via GitHub Copilot Chat)

Prerequisites

• Visucal Studio Code OR Visual Studio Code Insiders
• GitHub Copilot extension:
  • Make sure you have a GitHub Copilot license or subscription.
  • Install and enable the GitHub Copilot extension from the VS Code Marketplace.
  • GitHub Copilot Chat enabled in VS Code
  • For agent mode: VS Code Insiders with Custom Copilot Agent (Preview) support
• A transcript or note file, e.g., AIStoryMapper/inputs/yourfile.txt

Installation

1. Clone the Repository

   git clone https://github.com/UrjaSoni/AIStoryMapper
   
   cd AIStoryMapper

2. Open in VS Code

• Launch Visual Studio Code and open this repository’s folder.

Features

• Extracts detailed, ready-for-refinement Agile stories and spikes
• Parses raw transcripts and maps conversation into actionable work
• Supports metadata tagging (e.g., UX, DevX, AI, Reporting, etc.)
• Flags vague or unclear areas as refinement and marks complete items as ready
• Output is structured in markdown format and can be saved as a story map document

Use Cases

• Turn customer calls, discovery sessions, or brainstorm notes into backlog items
• Rapidly generate user stories, acceptance criteria, clarifying questions, and tags
• Standardize documentation for handoff to engineering teams

Usage

Option 1: Run with Custom Agent (VS Code Insiders Preview)

If you're using VS Code Insiders and have access to GitHub Copilot Custom Agents, follow these steps:

1. Open the transcript or notes file Open the .txt file you'd like to extract user stories from (e.g., AIStoryMapper/inputs/contoso_mock_meeting_transcript.txt)

2. Open Copilot Chat Open the Copilot Chat pane from the sidebar (or press Cmd+I / Ctrl+I if you have the shortcut enabled).

3. Choose Your Custom Agent

   Click the dropdown in the upper-right of the Copilot Chat window (next to the "Send" button).

   Select the userstories Agent from the list of available agents.

4. Choose Claude Sonnet 3.5 | 3.7 | 4 premium models

   Ask it:

   Please extract user stories and spikes from this transcript/notes file.

5. Output

   The agent will:

   • Parse the content

   • Generate structured user stories and spikes

   • Include full metadata, tags, status (ready or refinement)

   • Output the results as a markdown document you can save

6. Save If you like the changes the GitHub Copilot made, you can click on 'Keep' button to save the changes.

Option 2: Prompt-Based (GitHub Copilot Chat – Standard VS Code)

If you do not have access to Custom Agent Mode, you can use GitHub Copilot Chat directly:

1. Open your transcript file (e.g., AIStoryMapper/inputs/contoso_mock_meeting_transcript.txt)

2. In Copilot Chat, paste the following prompt:

   Please extract well-structured user stories and spikes from the currently open transcript file.
   
   At the very top of your output, add this metadata block:
   
   # User Stories and Spikes - Contoso Integration Review  
   **Generated from:** [transcript or note file name] 
   **Date:** [date the document is generated]  
   **Total Items:** [count of all user stories + spikes]  
   
   For each story or spike, include:  
   - Title  
   - User persona  
   - Goal (I want to...)  
   - Description  
   - Acceptance criteria  
   - Tags  
   - Clarifying questions/gaps  
   - Status (Ready or Refinement)
   
   Format the output as a numbered markdown list.
   
   Create all relevant items needed to fully cover the content of the transcript.
   
   Provide the output as if you are creating a new standalone markdown document that can be saved uniquely each time (e.g., user-stories-YYYYMMDD-HHMMSS.md).
   
   Do not add anything outside this format.
   
   

Future Improvements

• Shell script or CLI for auto-parsing and saving markdown output

• Optional integration with Jira, Linear, or Azure DevOps

• Integration with Loop or One-Note (if possible)