Bokata: Let LLM Help You Break Down Complex Requirements into Executable Development Roadmaps

Bokata is an open-source prompt engineering framework that leverages large language models (LLMs) to help convert vague product requirements into structured, incremental, and executable development roadmaps using vertical slicing and user story mapping techniques. It addresses pain points in traditional requirement breakdowns such as information overload, difficulty in grasping granularity, vague value orientation, and high communication costs, providing development teams with an efficient requirement analysis tool.

Traditional requirement analysis relies on manual work and has four main pain points: information overload (too many functional points in PRD, making it hard to identify priorities and dependencies), difficulty in grasping granularity (too detailed leads to loss of overall perspective; too coarse makes it hard to estimate schedule risks), vague value orientation (developing by module instead of user value), and high communication costs (product and development teams have understanding gaps that require repeated confirmation). Bokata's approach: Use LLM's natural language understanding and structured thinking capabilities to assist in requirement breakdown.

Bokata is an open-source prompt engineering framework specifically designed to guide LLMs in decomposing complex software requirements into incremental, high-value executable development plans. The name metaphorically refers to cutting complex requirements (big stones) into manageable small pieces (tasks). The project is hosted on GitHub at abrahamvallez/skills.

Vertical Slicing

Unlike horizontal layering (database first, then API, then frontend), vertical slicing requires each delivery unit to span all technical layers and provide end-to-end user value. Example: The first slice of an e-commerce application implements "browsing products and adding to cart", including frontend/backend/database. Benefits include having demonstrable results in iterations, early exposure of risks, and timely integration of user feedback.

User Story Mapping

A requirement organization method proposed by Jeff Patton, where the upper layer is user activities, the middle layer is tasks, and the lower layer is stories. It can show the full picture of the product, identify core MVP stories, and plan incremental releases. Bokata integrates both methods into prompt templates to guide LLMs in reorganizing requirements according to the methodology.

1. Input Phase: Provide original requirements (text description or PRD);
2. Analysis Phase: Prompt templates guide LLM to identify core user roles, activity flows, and task nodes;
3. Slicing Phase: Reorganize into end-to-end functional slices according to vertical slicing principles;
4. Mapping Phase: Organize into a user story map (horizontal user journey, vertical implementation plan);
5. Output Phase: Generate a structured development plan including priorities, dependencies, and schedule estimates.

Walking Skeleton, proposed by Alistair Cockburn, refers to the minimal feasible implementation that runs through the system architecture: only the basic skeleton structure, end-to-end workable, proving the feasibility of the core architecture, and subsequent development gradually enriches it. Difference from MVP: MVP emphasizes the minimal set of user value, while Walking Skeleton emphasizes the complete integration of the technical architecture. Combining the two allows early verification of technical solutions and user trials.

Applicable Scenarios: Startup teams for quick launch (generate roadmaps from vague ideas), large project refactoring (sort requirement priorities), agile transformation enterprises (understand vertical slicing and story mapping), remote collaboration teams (reduce understanding gaps). Limitations: The plan generated by LLM needs manual review and may miss business constraints; for highly innovative projects, traditional analysis is more appropriate.

Bokata encodes best practices in software engineering into prompt engineering, making LLM an assistant to requirement analysts. Its value lies in providing a structured thinking framework rather than replacing human judgment. Whether you use the framework directly or not, vertical slicing and user story mapping are reliable practices, and the idea of combining them with LLM shows more possibilities of prompt engineering in the development process. The project is open-source and can be tried on GitHub.