agent-email-inbox

Pass

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The documentation includes examples for installing official tools and scripts from well-known services, such as the Tailscale installation script (tailscale.com), ngrok, and Cloudflare's tunneling client (cloudflared). These resources are used for standard environment setup and local development tunneling.
  • [COMMAND_EXECUTION]: The skill provides standard command-line examples for package management (npm, pip), tool installation (brew), and service configuration (sudo tailscale). These commands are intended for administrative setup of the development environment.
  • [INDIRECT_PROMPT_INJECTION]: As the skill is designed to process untrusted inbound email, it specifically addresses the risk of indirect prompt injection. It provides detailed implementation patterns for multiple security levels, including strict sender allowlists, domain filtering, and content sanitization (stripping quoted threads and truncating content) to ensure untrusted data is handled safely by the agent.
  • [CREDENTIALS_UNSAFE]: The skill mentions sensitive environment variables like RESEND_API_KEY and RESEND_WEBHOOK_SECRET, but correctly instructs the user to manage these using secure environment files (.env) and provides clear warnings against sharing keys in conversation history.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 18, 2026, 03:43 AM