The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the local cortex CLI using the subprocess module. These calls are handled securely by passing arguments as a list rather than a shell string, which prevents command injection. The execute_cortex.py script further wraps this execution with a blocklist-based security model.
[SAFE]: Implements a comprehensive PromptSanitizer that automatically identifies and replaces personally identifiable information (PII) such as emails, phone numbers, and credit card patterns before they are processed.
[SAFE]: Includes a robust ApprovalHandler that predicts necessary tools and requires explicit user consent before execution in its default 'prompt' mode, ensuring human-in-the-loop oversight.
[SAFE]: Features a structured AuditLogger that records all execution events, routing decisions, and results to a local JSONL file with support for rotation and retention policies.
[SAFE]: Uses secure file handling practices, explicitly setting file and directory permissions to 0600/0700 to ensure that sensitive configurations and cached data are only accessible by the current user.

cortex-code