Control-Alt-Defeat

A multiple-choice exam app built with Flask. Pick an exam, answer questions, and get graded results with full history tracking. Exams are plain JSON files — drop one in and it just works.

Goes without saying that as this is a vibe coded experiment, I'd suggest only running it locally.

NOTE: Be mindful of NDA's should you contribute questions and exams. In my case I deliberately only gave AI agents information about the exams that the exam body publishes themselves (i.e sample questions, syllabus documentation) before getting the agents to generate exam JSON.

Quick Start

Docker (recommended):

docker compose up --build   # first run
docker compose up           # subsequent runs
docker compose down         # stop

The app will be available at http://localhost:5000.

Local development:

pip install -r requirements.txt
python app/app.py

Environment Variables

Variable	Default	Description
SECRET_KEY	dev value (warns on startup)	Flask session signing key — set this in production
HISTORY_DB	app/data/history.db	Path to the SQLite history database
GITHUB_REPO	RedcentricCyber/Control-Alt-Defeat	Repo slug used by the "report inaccuracy" button
CAD_PRODUCTION	off	Set to 1 when serving over HTTPS — enables Secure session cookies and HSTS

Included Exams

Exam	Question Pool	Difficulty	Source
CREST CCT Multichoice Preparation	1821	Hard	ChatGPT, Claude, Gemini
CyberScheme CSTM	108	Medium	ChatGPT
CREST CPSA Multichoice Preparation	493	Medium	Gemini

Creating Your Own Exams

Add a JSON file to app/exams/ following this format:

{
  "title": "My Exam",
  "description": "A short description shown on the exam card.",
  "difficulty": "MEDIUM",
  "pass_mark": 70,
  "time_limit": 30,
  "num_questions": 20,
  "reward": "FLAG{optional-reward-for-passing}",
  "based_on": "Optional — what the exam is based on",
  "created_by": "Optional — author name",
  "links": [
    { "text": "Study Guide", "url": "https://example.com" }
  ],
  "questions": [
    {
      "category": "Topic Area",
      "question": "What is the answer to this question?",
      "options": {
        "A": "Wrong answer",
        "B": "Wrong answer",
        "C": "Correct answer",
        "D": "Wrong answer"
      },
      "correct": "C",
      "explanation": "Optional explanation shown after grading.",
      "skill": "Optional skill value to allow for a subcategory under the main one, as used on CREST syllabus documentation"
    }
  ]
}

I found Gemini code studio to be best at creating an exam using something similar to the below prompt.

Create an XYZ exam .json file, using the below format

{
  "title": "My Exam",
  "description": "A short description shown on the exam card.",
  "difficulty": "MEDIUM",
  "pass_mark": 70,
  "time_limit": 30,
  "num_questions": 20,
  "reward": "FLAG{optional-reward-for-passing}",
  "based_on": "Optional — what the exam is based on",
  "created_by": "Optional — author name",
  "links": [
    { "text": "Study Guide", "url": "https://example.com" }
  ],
  "questions": [
    {
      "category": "Topic Area",
      "question": "What is the answer to this question?",
      "options": {
        "A": "Wrong answer",
        "B": "Wrong answer",
        "C": "Correct answer",
        "D": "Wrong answer"
      },
      "correct": "C",
      "explanation": "Optional explanation shown after grading.",
      "skill": "Optional skill value to allow for a subcategory under the main one, as used on CREST syllabus documentation"
    }
  ]
}

The XYZ exam syllabus is below

CATEGORY|SKILL

Here are some sample questions to judge the skill level required

EXAMPLE QUESTIONS

Also note:
* Please make sure explanations are helpful to candidates revision and help them understand why an answer is correct.
* I want at least 5 questions per skill
* Please ensure all the options (incorrect and correct) are similar length/verbosity
* Minimise repetition and similar questions

Key fields:

• time_limit — minutes; set to 0 for untimed
• num_questions — how many to serve per session (can be less than the total pool)
• category — questions are sampled proportionally across categories so every topic is represented
• correct — the key (A/B/C/D) of the right answer
• reward — optional text (e.g. voucher code, CTF flag) shown once to users who pass with default exam settings (question count and time limit unchanged)

If running with Docker, the exams directory is mounted as a volume — just drop the file in app/exams/ and refresh the page. No rebuild needed.

Reporting Inaccuracies

Since the included exams were generated by AI, inaccuracies are possible. Use the Report button on any question during an exam to create a pre-filled GitHub issue, or create an issue directly with details of the problem and the correction.