Obfuscated Commands Audit Fixture
Provides deliberately flawed command patterns to test and validate skill security auditing tools.
Obfuscated Commands Skill is a synthetic test fixture from the ai-dev-toolkit repository designed specifically for benchmarking automated skill auditing tools such as audit-skills. By deliberately embedding high-risk patterns—including unwarranted sudo privilege escalation, recursive deletions reaching into user system directories, and base64-encoded command execution via eval—this fixture enables developers building security analyzers to ensure their scanners accurately detect dangerous, hidden, or out-of-scope behaviors in Claude Code skills.
主な機能
011 GitHub stars
02Tests static detection of base64-encoded and evaluated shell commands
03Simulates unauthorized sudo privilege escalation within project scripts
04Embeds out-of-scope recursive deletion patterns targeting user home directories
05Provides deterministic test fixtures for skill auditing rule verification
06Assists in benchmarking AI coding agent safety and permission filters
ユースケース
01Validating CI/CD test suites for Claude Code skill repositories
02Verifying that skill auditing tools correctly flag obfuscated shell payloads
03Benchmarking static security linters against dangerous privilege escalation patterns