sphinx-lookup 0.2.0

Lightweight CLI for querying the Sphinx Scanner LLM Attack Database

Sphinx Lookup

Lightweight CLI for querying the Sphinx Scanner LLM Attack Database

Query 1900+ documented LLM/AI attacks directly from your terminal. No Docker, Redis, or Elasticsearch required — just 4 dependencies.

Installation

pip install sphinx-lookup

Quick Start

# Configure your Supabase credentials (one-time setup)
sphinx-lookup configure

# Search for attacks
sphinx-lookup search "prompt injection"

# List attacks by category
sphinx-lookup list --category jailbreak

# Show full attack details
sphinx-lookup show <attack-id>

# View statistics
sphinx-lookup stats

Commands

Command	Description
configure	Interactive setup for Supabase credentials
search <query>	Search attacks by name or description
list	List attacks with optional filters
show <id>	Display full attack card (Steckbrief)
categories	Show category breakdown with counts
stats	Database overview statistics

Attack Categories

• prompt_injection - Manipulating LLM behavior through crafted prompts
• jailbreak - Bypassing safety filters and restrictions
• adversarial_input - Inputs designed to cause unexpected behavior
• data_extraction - Extracting training data or sensitive information
• model_manipulation - Altering model weights or behavior
• denial_of_service - Causing service disruption
• privacy_attack - Violating user privacy
• agent_manipulation - Targeting AI agents and tool use

Modalities

Filter attacks by input type: text, reasoning, vision, audio, multimodal, code

sphinx-lookup list --modality vision

Configuration

Credentials are read in this order (later overrides earlier):

1. .env in current directory
2. ~/.sphinx-lookup.env
3. Environment variables (SUPABASE_URL, SUPABASE_KEY)

Requirements

• Python 3.10+
• Supabase project with Sphinx Scanner database

Part of Sphinx Scanner

This is the lightweight lookup CLI from the Sphinx Scanner project. For the full crawling and extraction pipeline, see the main project.

License

MIT