google-forms
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION] (HIGH): The tool
get_form_responsescreates a significant indirect prompt injection surface by ingesting untrusted input from external form submitters. Because the agent also possesses write and delete capabilities, a malicious form submission could potentially manipulate the agent into performing unauthorized actions. 1. Ingestion points:get_form_responses(SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory:delete_form,delete_question,update_form(SKILL.md). 4. Sanitization: None specified. - [CREDENTIALS_UNSAFE] (HIGH): The setup instructions require users to store highly sensitive
GOOGLE_CLIENT_SECRETandGOOGLE_REFRESH_TOKENvalues in a plaintext.mcp.jsonfile. This exposes long-lived credentials to any user or process with read access to the local filesystem. - [EXTERNAL_DOWNLOADS] (MEDIUM): The
bun installcommand downloads third-party Node.js packages at setup time. These dependencies are not pinned to specific versions or verified via hashes in the instructions, posing a supply chain risk.
Recommendations
- AI detected serious security threats
Audit Metadata