code-lab-assist

A powerful academic programming tool that generates lessons and structured labs across multiple programming languages and environments. It's designed to streamline lessons for instructors and tutors with automated testing and vulnerability scanning. It also supports heuristic POML generation.

Features

• Multi-Language Support: Generate code templates for Python, JavaScript, TypeScript, Java, C, C++, C#, and SQL
• Versatile Code Templates: Generate functions, classes, and interfaces with appropriate language-specific syntax
• Automated Testing: Generate comprehensive test suites for all supported languages and environments
• Vulnerability Scanning: Playwright tests can run BrainScan directly in the browser runtime to detect risky patterns in all lab code
• User-Friendly Interface: Interactive Jupyter notebook interface for easy lab creation
• Heuristic POML Generation: Generate powerful and intuitive POML prompts for any purpose

Getting Started

Prerequisites

• Python 3.6+
• IDE support for Python, Jupyter notebooks, and GitHub Copilot (or an OpenCode installation instead of GitHub Copilot)
• requirements.txt dependencies installed (pip install -r requirements.txt)
• Language-specific compilers/interpreters for testing code
• A web browser, internet connection, and way to serve websites locally for the SQL Lab (e.g., Live Server extension in VS Code)
• Git access to the brainscan repository (required for vulnerability scanning)

Usage

1. Launch the Jupyter notebook interface (code_lab_ui.ipynb)
2. Use the interactive form to specify:
   • Programming language
   • Code type (function, class, or interface)
   • Name
   • Problem description
   • Parameters
3. Click "Generate" to create your lab

POML Renderer

The project includes a simplified POML renderer (poml_renderer.py) for rendering POML prompts with context variables. The renderer is designed to be minimal and straightforward:

python poml_renderer.py

Features:

• Renders prompt.poml
• Loads context from prompt_context.json (optional)
• Outputs rendered content directly to stdout
• Minimal, clean implementation without command-line complexity

Config Files:

• prompt.poml - The POML prompt file to render
• prompt_context.json - Context variables (optional, uses empty context if not found)

Browser Runtime Security Testing (BrainScan + Playwright)

BrainScan Submodule Setup

This project uses the brainscan repository as a git submodule for browser-based vulnerability scanning.

To clone the repository with the submodule:

git clone --recurse-submodules https://github.com/jesse318s/brainscan.git

Running the Playwright-based Vulnerability Scan Tests

The Playwright suite includes runtime security tests in playwright-tests/test_vulns.py. These tests inject the vendored BrainScan browser bundle (brainscan/vendor/brain.js/browser.js) and lab index src code into the page, restore the trained network snapshot from brainscan/data/trained-network.json, and scan runtime page content for vulnerability indicators.

Key behavior:

• Scanning happens inside the browser runtime (not through a separate scan API call)
• The injected BrainScan script is excluded from the scanned content to avoid false positives from the scanner implementation itself
• Tests fail when the predicted risk reaches the Critical threshold

If the trained snapshot file does not exist yet, generate it by running BrainScan once from brainscan/:

cd brainscan && node index.js && cd ..

This will create brainscan/data/trained-network.json (gitignored) for use by the tests. (It may need to be recreated if it causes false positives)