HardStop acts as a security sentinel for the agent's shell interactions. It intercepts commands to check for dangerous patterns (like rm -rf /, pipe-to-shell patterns, or unsafe sudo usage) and blocks access to sensitive credential files (e.g., .env, .ssh).
This skill should be active for every shell command or file read operation, especially when the agent is executing suggestions from external sources like Stack Overflow or performing sysadmin tasks.
Compatible with any agent with shell/filesystem access (Claude Code, Cursor, etc.).
This skill has not been reviewed by our automated audit pipeline yet.