The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill provides patterns for executing AppleScript via osascript. The provided code includes security controls like timeouts, process management, and error handling in references/advanced-patterns.md.
[DATA_EXFILTRATION] (SAFE): While identifying exfiltration as a risk, the code includes defensive measures like shell command allowlisting and blocking network-capable commands in references/security-examples.md.
[PROMPT_INJECTION] (LOW): The skill includes patterns for processing untrusted data, identifying an indirect injection surface (Category 8). 1. Ingestion points: params dictionary in ScriptTemplates.render and build_safe_script. 2. Boundary markers: Brace interpolation ({key}) without robust delimiters. 3. Capability inventory: Execution of generated scripts via osascript using subprocess and asyncio. 4. Sanitization: Manual escaping of backslashes and double quotes, plus regex-based application name validation.
[SAFE] (SAFE): The automated scan alert for self.logger.info is a false positive; the code implements auditing for script executions, which is a recommended security control.

applescript