resend
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Local Command Execution: The skill utilizes
npxandcurlto manage a local emulation service. These tools are used to start the emulator and interact with its REST API athttp://localhost:4000. This is standard practice for local development workflows. - Indirect Prompt Injection Surface: The skill is designed to process and store email content, including HTML and text. This creates a potential surface where external data (simulated emails) could contain instructions that an agent might encounter while performing tasks like verification code extraction.
- Ingestion points: Untrusted data enters the context via
POST /emailsandPOST /emails/batchrequests (documented inSKILL.md), specifically through fields likehtml,text, andsubject. - Boundary markers: The skill uses standard JSON structures for data handling, though it does not explicitly define prompt-level delimiters for the email content itself.
- Capability inventory: The skill utilizes
Bashforcurlandnpxoperations to manage the local server and query data. - Sanitization: The documentation does not specify sanitization for the content of the simulated emails, as it is intended for high-fidelity emulation of a production API.
Audit Metadata