storage-analyzer

Warn

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses subprocess.run to execute disk usage analysis tools (du) and system information utilities (diskutil, uname, sw_vers) within the scripts/scan.py script.\n- [COMMAND_EXECUTION]: The scripts/server.py script provides a local HTTP server that exposes functionality to delete files (rm) or move them to the trash (trash) via POST requests.\n- [COMMAND_EXECUTION]: On macOS, the skill utilizes osascript to perform file deletions through the Finder application, and on Windows, it uses ctypes to call SHFileOperationW for recycling files.\n- [COMMAND_EXECUTION]: The skill executes shell commands to open directories in the host system's file manager (open on macOS, explorer on Windows) via scripts/server.py.\n- [DATA_EXFILTRATION]: The scripts/scan.py script recursively scans the user's home directory and application data folders, collecting detailed metadata about file paths and sizes to be processed by the AI agent.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes unvalidated file names and paths from the user's file system (ingestion points in scripts/scan.py) and uses them in classification logic that drives file deletion capabilities (capability inventory in scripts/server.py). There are no boundary markers to prevent malicious instructions in file names from influencing the agent's behavior.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 12, 2026, 09:16 PM
Security Audit — agent-trust-hub — storage-analyzer