google-drive

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (Google Drive file names, descriptions, and comments) that could contain malicious instructions designed to manipulate the agent.\n
  • Ingestion points: Data is fetched in scripts/google-drive.py via functions like list_files, search_files, get_file_metadata, and list_comments.\n
  • Boundary markers: Absent. Drive content is not delimited or specifically marked as untrusted data before being presented to the agent context.\n
  • Capability inventory: The script scripts/google-drive.py has the ability to delete files (delete_file), change permissions (share_file), and upload new content, which could be abused if an injection succeeds.\n
  • Sanitization: Absent. Data retrieved from the API is formatted and returned to the agent without escaping or structural validation.\n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of several well-known Python packages from PyPI during setup.\n
  • Packages: google-auth, google-auth-oauthlib, google-api-python-client, keyring, and pyyaml.\n- [COMMAND_EXECUTION]: The skill uses a Python script (scripts/google-drive.py) as the primary mechanism for interacting with the Google Drive API.\n- [DATA_EXFILTRATION]: The skill manages access to sensitive personal data and authentication tokens.\n
  • Evidence: It correctly implements token management by using the system keyring and avoiding hardcoded credentials. It requires the user to perform an interactive OAuth flow, ensuring human-in-the-loop authorization.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 11:07 AM
Security Audit — agent-trust-hub — google-drive