The Agent Skills Directory

COMMAND_EXECUTION (MEDIUM): Path traversal vulnerability in scripts/drive_transfer.py via the dest_folder argument. Evidence: The run_drive_transfer function uses os.path.join to construct the destination path without sanitizing for parent directory references (..). This allows a user or malicious data source to write files to arbitrary locations on the host system.
COMMAND_EXECUTION (SAFE): Robust security implementation in scripts/transfer.py. Evidence: This module includes sanitize_path for traversal detection, a strict regex for filename validation (^[a-zA-Z0-9_.-]+$), and a whitelist of allowed Android destination directories. It correctly uses shlex.quote for all shell-command parameters.
DATA_EXPOSURE & EXFILTRATION (SAFE): System maintenance activities are restricted to expected directories. Evidence: Cleanup operations are targeted at ~/.Trash, ~/Library/Caches, and ~/Downloads. The skill uses safe standard library functions and includes a dry-run mode to prevent accidental data loss.
INDIRECT PROMPT INJECTION (LOW): Vulnerability surface identified. 1. Ingestion points: source and dest_folder command-line arguments. 2. Boundary markers: Absent. 3. Capability inventory: File system write/delete and ADB interaction. 4. Sanitization: Comprehensive for the transfer command, but insufficient for drive-transfer.

MacPowerTools