Claude Skill
XSafeAI/XSafeClaw
XSafeClaw is an open-source agent safety platform from Fudan University for evaluating LLM-based agent security against prompt injection and red-teaming threats.
Overview
Stars150
Forks6
LanguagePython
Last pushed2026-05-27
Last synced2026-05-27
Repository
OwnerXSafeAI
RepositoryXSafeClaw
Full nameXSafeAI/XSafeClaw
Repo ID1,177,710,325
GitHub URLhttps://github.com/XSafeAI/XSafeClaw
Install this Skill
git clone https://github.com/XSafeAI/XSafeClaw.gitRegistry
TypeUnknown
Quality scoreUnknown
VerificationUnknown
Last verifiedUnknown
Summary
XSafeClaw is an open-source agent safety platform developed by Fudan University, designed to evaluate and enhance the security of LLM-based agents against prompt injection, red-teaming, and other AI safety threats.
Chinese description
介绍XSafeClaw：复旦大学开源智能体安全平台
Key features
- Open-source agent safety evaluation platform from Fudan University
- Supports prompt injection and red-teaming testing
- Focuses on LLM-based agent security
- Built for AI safety research and benchmarking
Use cases
- Security testing of LLM-based agents
- Red-teaming exercises for AI systems
- Academic research on agent safety
- Benchmarking agentic AI defenses