Claude Skill
trustunknown/thomas
TrustedExecBench is an open-source TypeScript framework for scenario-grounded security evaluation and red-teaming of autonomous personal AI assistants.
Overview
Stars343
Forks51
LanguageTypeScript
Last pushed2026-05-08
Last synced2026-05-15
Repository
Ownertrustunknown
Repositorythomas
Full nametrustunknown/thomas
Repo ID1,081,166,515
GitHub URLhttps://github.com/trustunknown/thomas
🚀 Install this Skill
openclaw install trustunknown/thomasSummary
TrustedExecBench is a scenario-grounded security evaluation framework for autonomous personal AI assistants. It provides a structured benchmark to assess and red-team the security of AI agents in realistic execution environments.
Chinese description
TrustedExecBench：面向自主个人AI助手的场景化安全评估。
Key features
- Scenario-grounded security evaluation for AI assistants
- Built-in red-teaming capabilities for agent security
- Structured benchmark with realistic execution environments
- Open-source TypeScript framework for reproducible testing
- Focus on autonomous personal AI assistant security
Use cases
- Security evaluation of autonomous AI assistants
- Red-teaming AI agents for vulnerability discovery
- Benchmarking agent security in realistic scenarios
- Research on AI assistant safety and trustworthiness
- Developing secure personal AI assistant systems