Visit this page to download and run the app on Windows:
- Open the link above in your browser.
- On the GitHub page, look for the latest release or the main download files.
- Download the Windows file if one is listed.
- If the file comes in a ZIP folder, right-click it and choose Extract All.
- Open the extracted folder.
- Double-click the app file to run it.
- If Windows asks for permission, choose Yes.
This project helps you work with AI coding tools in a clear order.
It guides you through:
- writing a short spec
- making a plan
- breaking work into tasks
- building the change
- checking the result
This helps you keep work organized when you use tools like Claude Code, Cursor, GitHub Copilot, and similar apps.
When you start with a clear spec, you spend less time fixing confusion later.
This skill helps you:
- keep the goal clear
- avoid missed details
- reduce rework
- compare the plan with the result
- spot drift between the request and the code
It fits common software work such as:
- new features
- bug fixes
- UI changes
- product updates
- small refactors
For Windows use, you only need:
- a Windows PC
- an internet connection
- a web browser
- enough free space to download the file
- permission to run downloaded apps
If you plan to use this with an AI coding tool, you may also want:
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- another editor or coding agent
This repository is built around a spec-first workflow.
Typical parts include:
- a clear product spec
- a plan based on that spec
- a task list
- implementation steps
- validation checks
- drift checks to keep work on track
Write what you want in plain language.
Good specs answer:
- what the user needs
- why it matters
- what the result should do
- what is out of scope
Turn the spec into a simple plan.
The plan should cover:
- main steps
- file or area changes
- risks
- edge cases
- needed checks
Break the plan into small jobs.
Good tasks are:
- short
- clear
- easy to test
- in order
Use your AI coding tool to make the change.
Keep the tool focused on:
- one task at a time
- the current plan
- the spec details
Check that the result matches the spec.
Use checks such as:
- manual testing
- simple acceptance criteria
- code review
- drift detection
This skill works well when you want to:
- build a feature from a clear request
- keep AI output tied to a goal
- avoid vague coding prompts
- manage complex work in steps
- track changes from spec to code
It is also helpful for:
- product teams
- solo builders
- support fixes
- prototype work
- ongoing feature updates
You want a new settings page.
Use this order:
- write the request
- define what the page must do
- list the tasks
- build one task at a time
- check each result against the spec
That gives the AI a clear path and gives you a clear review path
Validation means checking if the result matches the request.
Drift detection means checking if the work has moved away from the spec.
Look for:
- missing steps
- extra changes
- changed behavior
- skipped acceptance criteria
- code that solves a different problem
This keeps the work aligned from start to finish
Use short, direct prompts.
Try to:
- name the goal
- state the expected result
- list limits
- include acceptance criteria
- ask for one step at a time
Avoid broad prompts like:
- make this better
- fix the app
- improve everything
Those prompts give the agent too much room to guess
This workflow fits common AI coding tools and editors, including:
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- similar agent-based coding tools
The steps stay the same even if the tool changes
- Download the project from the link above.
- Open it in your browser or file manager.
- Read the spec files or workflow notes.
- Use the steps in order: specify, plan, tasks, implement, validate.
- Keep each change small.
- Compare the result with the original request.
Before opening any downloaded file:
- check the file name
- make sure it comes from the GitHub link above
- keep your antivirus on
- use the latest Windows updates
- avoid unknown files from other sites
You may see files such as:
.zipfor a packed folder.mdfor markdown notes.jsonfor structured data.txtfor plain text.exefor a Windows app
If you see a ZIP file, extract it first before opening the contents
- Open the download link.
- Get the latest Windows file or project files.
- Extract the files if needed.
- Open the folder.
- Read the spec and workflow files.
- Follow the steps in order.
When you open the project, look for files that describe:
- the goal
- the plan
- the task list
- checks to run
- rules for the agent
These files help the AI stay focused and help you review the result with less guesswork
- Name: spec-driven-development-skill
- Type: Agent skill for spec-driven development
- Focus: specification-first workflow
- Tools: Claude Code, Cursor, GitHub Copilot, and similar tools
- Topics: acceptance criteria, requirements, design doc, drift detection, prompt engineering, and software engineering
You know the workflow is working when:
- the spec is clear
- the plan matches the goal
- tasks are small and ordered
- the code changes stay on target
- the final result matches the acceptance criteria
If you adapt this skill for a project, keep this order:
- Spec
- Plan
- Tasks
- Build
- Check
That structure keeps the work easy to follow and easy to review
If Windows does not open the file:
- check that it finished downloading
- right-click the file and try again
- extract ZIP files first
- confirm you are using the correct app for the file type
- reopen the GitHub page and get the file again