The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The list_directory.py script accepts a path argument without any sanitization or sandboxing, creating a reconnaissance surface. * Ingestion points: The script reads from sys.argv[1] in scripts/list_directory.py. * Boundary markers: Absent. There are no restrictions to keep the listing within a specific workspace or directory. * Capability inventory: Provides the ability to list all files, check directory status, and retrieve file sizes across any accessible path on the host via os.listdir. * Sanitization: None. The script directly passes the input string to os.listdir() and os.path.abspath().
[Data Exposure] (LOW): get_system_info.py gathers system fingerprinting data. * Evidence: The script collects platform.node() (hostname), platform.version(), and platform.processor(), which can be used to identify specific host environments and potential vulnerabilities.

system-info