recipe-draft-email-from-doc
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
gwsCLI commands to retrieve document data and send emails. This tool is a vendor-specific resource for the Google Workspace author. - [DATA_EXFILTRATION]: The skill implements a data flow where information is read from Google Docs and transmitted via Gmail. This behavior matches the primary stated purpose of the skill and occurs within the vendor's ecosystem.
- [PROMPT_INJECTION]: The skill is structurally vulnerable to indirect prompt injection because it processes untrusted data from external documents.
- Ingestion points: Google Doc content is retrieved in Step 1 of the instructions.
- Boundary markers: Absent. The content is interpolated directly into the email body without delimiters.
- Capability inventory: The skill possesses the capability to send outgoing network communications via the Gmail tool.
- Sanitization: No sanitization or safety instructions are included to prevent the agent from obeying instructions embedded within the document content.
Audit Metadata