noaiguarde 0.1.0

A lightweight Python package for first-level prompt safety checking using substring matching

noaiguarde 🛡️

A lightweight Python package for first-level prompt safety checking using simple substring matching.

Installation

pip install noaiguarde

Or with uv:

uv add noaiguarde

Quick Start

from noaiguarde import check_prompt

# Safe prompts return True
check_prompt("Hello, how are you?")  # True
check_prompt("What's the weather today?")  # True

# Unsafe prompts return False
check_prompt("I hate you")  # False
check_prompt("Let's attack the server")  # False

How It Works

noaiguarde performs case-insensitive substring matching against a bundled word list. If any word from the list is found within the prompt, it returns False (unsafe). Otherwise, it returns True (safe).

This is designed as a first-level filter - a quick, lightweight check before sending prompts to more sophisticated AI moderation systems.

API Reference

check_prompt(prompt: str) -> bool

Check if a prompt is safe.

Parameters:

• prompt (str): The text prompt to check.

Returns:

• True if the prompt is safe (no bad words found).
• False if the prompt contains any bad words.

Example:

from noaiguarde import check_prompt

if check_prompt(user_input):
    # Safe to proceed
    send_to_ai(user_input)
else:
    # Block or flag the input
    handle_unsafe_input(user_input)

Use Cases

• 🔒 Pre-flight checks before sending prompts to AI APIs
• 🚫 Content filtering in chatbots and user interfaces
• ⚡ Quick screening before expensive AI moderation calls
• 🛡️ Defense in depth as part of a multi-layer safety strategy

Limitations

• Substring matching only: May produce false positives (e.g., "skill" contains "kill")
• English only: The bundled word list is in English
• Not a complete solution: Use alongside more sophisticated AI safety tools

License

MIT License - see LICENSE for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.