Skip to content

Speckit

link from github

first setup the template for your AI Coding Assistant for your project

providing prompts to speckit

NOTE: * used the model "SWE-1.5" from Windsurf Cascade to generate speckit documentation * used the model "Claude Opus 4.5" to generate code

/speckit.constitution

  • the application should adhere to SOLID design principles. it should have a test coverage of 90%.
  • it should have a modular structure with backend services and user interface client calling these services
  • the application should show usage of at minimum 15 of the below
    • arrays
    • user defined objects
    • objects as data records
    • simple selection (if else)
    • complex selection (nested if, if with multiple conditions or switch)
    • loops
    • nested loops
    • user defined methods
    • user defined methods with parameters
    • user defined methods with return values
    • sorting
    • searching
    • file input output
    • use of sentinels or flags
    • recursion
    • merging of two or more sorted data structures
    • polymorphism
    • inheritance
    • encapsulation
    • parsing a text file

/speckit.specify

  • this is a standalone application
  • use a pre-defined list of questions and answers stored in a text/csv file
  • user is aked what topic he wants from a pre-defined list;
  • user is also asked a difficulty level -- choose from 3, easy-medium-hard
  • application randomly picks one question from the topic for that difficulty level
  • application receives answer from user, checks against the dataset (question-bank)
  • application also keeps track of user's scores
  • at the end, the application summarizes the results

/speckit.clarify

  • run this slash-command
  • let the AI Code Assistant analyze and ask clarifying questions
  • answer the questions with your option

/speckit.plan

  • implement this application in python 3.12 using UV from Astral as package manager.
  • make use of panda library to read list of Question objects from the CSV
  • the CSV header specifies the Question Class
  • use FastAPI to create backend services for this application
  • create two user interfaces, one should be a Command line interface
  • second user interface should be in Jinja2 using HTMX

/speckit.tasks

  • just run the slash-command

/speckit.implement

Ask AI assistant to implement only certain tasks at time
Or you can ask it to implement phase by phase, where phases are a collection of tasks in a defined sequence.