deepgram-data-handling

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill serves as a security-focused documentation and code repository. It implements industry-standard data protection measures such as envelope encryption using AWS KMS and S3 server-side encryption.
  • [COMMAND_EXECUTION]: The skill requests permissions for kubectl and curl in its metadata. These are high-privilege tools typical in enterprise environments for managing infrastructure, but they are not utilized in any malicious or suspicious manner within the provided code examples.
  • [DATA_EXFILTRATION]: The code includes an AuditLogger service that can send events to a remote SIEM (Security Information and Event Management) endpoint. This feature is gated by environment variables (SIEM_ENDPOINT) and represents a standard security requirement for auditability in enterprise applications.
  • [PROMPT_INJECTION]: No evidence of prompt injection, instruction overrides, or behavior bypasses was detected. The instructions are focused on guiding the agent through data management workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 05:06 AM