Designed specifically for testing automated Claude Code skill scanners such as skillspector-scan, this test fixture simulates an environment variable diagnostics utility while intentionally incorporating high- and critical-severity detection patterns like credential harvesting, network exfiltration, and dynamic execution. It serves as a controlled benchmark to prove that automated security scanners and CI/CD exit-code gates reliably trigger DO_NOT_INSTALL blocks when evaluating untrusted or compromised AI agent skills.
主要功能
01Simulates environment diagnostics to test static and dynamic analysis tools
02Validates CI/CD exit-code gating and automated policy enforcement
03Tests scanner resilience against credential harvesting and dynamic execution
041 GitHub stars
05Provides a reproducible test case for AI agent skill security benchmarks
06Contains unambiguous high- and critical-severity security detection patterns
使用场景
01Benchmarking static analysis and LLM-based security linters for AI skills
02Verifying that Claude Code skill security scanners reliably flag critical threats
03Testing CI/CD exit-code gates designed to block risky agent capabilities