LostOxygen/llm-confidentiality
Whispers in the Machine: Confidentiality in Agentic Systems
This project helps evaluate how securely large language model (LLM) agents handle sensitive information when they use external tools like calendars or email. It takes an LLM agent setup, defines a 'secret' the agent must protect, and then tests its vulnerability to various malicious instructions. The output shows how easily the agent might accidentally reveal that secret, making it useful for developers, AI security researchers, or anyone deploying LLM agents to understand and mitigate data leakage risks.
Use this if you are building or deploying LLM-based agents that interact with external services and need to understand their susceptibility to prompt injection attacks that could leak confidential data.
Not ideal if you are looking for a pre-packaged defense solution or a tool to protect static data without an agentic system involved.
Stars
42
Forks
6
Language
Python
License
Apache-2.0
Category
Last pushed
Dec 11, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/prompt-engineering/LostOxygen/llm-confidentiality"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
protectai/llm-guard
The Security Toolkit for LLM Interactions
MaxMLang/pytector
Easy to use LLM Prompt Injection Detection / Detector Python Package with support for local...
utkusen/promptmap
a security scanner for custom LLM applications
agencyenterprise/PromptInject
PromptInject is a framework that assembles prompts in a modular fashion to provide a...
Resk-Security/Resk-LLM
Resk is a robust Python library designed to enhance security and manage context when...