obscuralabs-AI/Symbolic-Prompt-PenTest

Semantic Stealth Attacks & Symbolic Prompt Red Teaming on GPT and other LLMs.

21
/ 100
Experimental

This project helps AI security professionals and researchers evaluate the robustness of large language models (LLMs) against sophisticated, non-obvious attacks. It provides a method for crafting 'symbolic prompts'—metaphorical narratives that can bypass standard defenses. You input these stealthy prompts into an LLM and analyze its responses to understand vulnerabilities, using the provided scoring system to measure the model's resistance.

No commits in the last 6 months.

Use this if you are responsible for testing the security and ethical alignment of LLMs and need advanced techniques to uncover hidden vulnerabilities that keyword-based red teaming misses.

Not ideal if you are looking for traditional software penetration testing tools or simple keyword-based prompt injection techniques.

AI Security LLM Red Teaming Model Evaluation Prompt Engineering AI Safety
No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 4 / 25
Maturity 7 / 25
Community 8 / 25

How are scores calculated?

Stars

8

Forks

1

Language

License

Last pushed

May 16, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/prompt-engineering/obscuralabs-AI/Symbolic-Prompt-PenTest"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.