Skip to content

Geboler/slice-based-project-development-skill

Repository files navigation

Slice-Based Project Development Skill

A reusable workflow skill for managing AI-assisted software projects through small, testable, acceptance-driven development slices.

This repository is designed for long-running development projects where a user works with coding agents such as Codex, ChatGPT, Claude, or other LLM-based development assistants.

What This Skill Does

This skill helps you:

  • Break a complex project into small development slices
  • Generate structured development prompts
  • Generate independent acceptance prompts
  • Track slice status
  • Produce handoff summaries between sessions
  • Decide the next slice after acceptance
  • Avoid context overload in long-running projects
  • Prevent unrelated refactors and scope drift

Core Idea

Each slice should have:

  1. A clear goal
  2. A limited scope
  3. Explicit out-of-scope boundaries
  4. Testable acceptance criteria
  5. A development prompt
  6. An acceptance prompt
  7. A handoff summary
  8. A recommended next slice

Recommended Use Cases

  • AI-assisted software development
  • Codex project iteration
  • Web app MVP development
  • Long-running feature development
  • Multi-session project handoff
  • Acceptance-driven development
  • Complex project planning with many small features

Repository Structure

slice-based-project-development-skill/
├── README.md
├── SKILL.md
├── project_profile.template.md
├── CHANGELOG.md
├── extensions/
│   └── superpower_layer.md
├── workflows/
│   ├── 01_slice_planning.md
│   ├── 02_development_prompt.md
│   ├── 03_acceptance_prompt.md
│   ├── 04_handoff_summary.md
│   └── 05_next_slice_decision.md
├── templates/
│   ├── development_prompt.template.md
│   ├── acceptance_prompt.template.md
│   ├── handoff_summary.template.md
│   ├── issue_report.template.md
│   └── slice_registry.template.md
└── examples/
    └── generic_web_app/
        ├── project_profile.md
        ├── slice_034_delete_record.md
        ├── slice_034b_delete_confirmation_ui.md
        ├── slice_034c_acceptance.md
        ├── slice_035_acceptance_prompt.md
        └── slice_035_resource_slot_binding_prep.md

How to Use

  1. Copy this repository into your workspace or keep it as a standalone GitHub repository.
  2. For each project, copy project_profile.template.md and fill in project-specific details.
  3. Before starting a feature, create a new slice using the workflow in workflows/01_slice_planning.md.
  4. Use templates/development_prompt.template.md to generate a development prompt for Codex or another coding agent.
  5. Use templates/acceptance_prompt.template.md to generate an independent acceptance prompt.
  6. After acceptance, write a handoff summary using templates/handoff_summary.template.md.
  7. Decide the next slice using workflows/05_next_slice_decision.md.

Session Roles

When publishing a task to an AI coding agent, state the session role explicitly. The default roles are planning, development, acceptance, handoff, and repair, but each project can define its own role map.

For example, a project may split work into integration, web/app, backend, integrated acceptance, renderer, or other project-specific sessions. Record those roles in the project profile so each prompt has clear ownership, boundaries, and handoff expectations.

Optional Superpower Layer

Use extensions/superpower_layer.md when a project needs stronger engineering discipline around scope control, acceptance quality, repeated slices, or cross-slice consistency.

The Superpower Layer is an optional enhancement layer. It does not replace the core slice workflow or change the standard planning, development, acceptance, handoff, and next-slice decision steps.

Suggested Repository Visibility

Use a public repository if this skill only contains general workflow rules and sanitized examples.

Use a private repository if the examples contain:

  • Internal project architecture
  • Private business logic
  • API endpoints
  • Database schema
  • Commercial product details
  • User data or sensitive configuration

License

You may add your preferred license later. If unsure, start without a license or use MIT for public reusable workflow material.

About

A reusable workflow skill for managing AI-assisted software projects through small, testable, acceptance-driven development slices.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors