google-drive-automation

Warn

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: Instructions require running local Python scripts (scripts/auth.py and scripts/drive.py) that were not included in the provided skill package, making their behavior unverifiable.
  • [DATA_EXFILTRATION]: The skill documentation states that it refreshes expired tokens using Google's cloud function. This indicates a non-standard OAuth implementation where tokens might be transmitted to a remote endpoint managed by the skill author rather than official Google OAuth endpoints.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from Google Drive (file names, content search, and downloads), which could contain malicious instructions designed to influence the agent's behavior.
  • Ingestion points: File data and metadata retrieved via drive.py search, list, and download commands in SKILL.md.
  • Boundary markers: None provided in instructions or script references to delimit external content.
  • Capability inventory: Subprocess execution (python scripts/...), file system writes (download to ./downloads/), and network access (implied Drive API operations).
  • Sanitization: None evident for data retrieved from external Drive files before it is processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 6, 2026, 06:26 AM
Security Audit — agent-trust-hub — google-drive-automation