gmail-label

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted email snippets and subjects through an LLM-based subagent for classification. A malicious email could contain instructions designed to manipulate the classification process.\n
  • Ingestion points: scripts/gmail_label_fetch.py retrieves email metadata (Subject, Snippet) from the Gmail API, which is then stored in .tmp/emails.json.\n
  • Boundary markers: No specific boundary markers or 'ignore' instructions are used in the subagent prompt defined in SKILL.md to isolate email content from classification instructions.\n
  • Capability inventory: The skill possesses the ability to modify Gmail labels and execute local Python scripts via the Bash tool.\n
  • Sanitization: scripts/gmail_label_apply.py performs hex-string validation on message IDs to prevent command injection, but does not sanitize the content being classified.\n- [SAFE]: The skill interacts with the Gmail API to fetch and modify email labels using official Google client libraries, which is consistent with its stated purpose.\n- [SAFE]: The skill manages authentication via local configuration files (gmail_accounts.json) and OAuth token files as part of its documented workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 12:42 AM
Security Audit — agent-trust-hub — gmail-label