gmail-workflows
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted external inputs from incoming emails, exposing it to potential indirect prompt injection vulnerabilities.
- Ingestion points: Processes email parameters such as sender, subject content, and text from attachments in workflows defined in
SKILL.md. - Boundary markers: Absent; there are no clear definitions or prompt boundaries isolating untrusted email metadata/content from workflow instructions.
- Capability inventory: Employs MCP tools for searching Gmail, extracting attachments, modifying email states (labeling/archiving), and writing data/files to Google Drive.
- Sanitization: Does not document any content escaping, schema enforcement, or validation logic prior to processing the ingested email streams.
Audit Metadata