gpt-image-2.5
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements strict validation for image inputs, checking both file extensions and magic bytes for PNG, JPEG, and WEBP formats to ensure file integrity.\n- [SAFE]: Network operations are secured with specific timeouts and file size limits (50MB for images, 4MB for masks), preventing resource exhaustion during remote image downloads.\n- [SAFE]: Credential management follows best practices by retrieving the Frevana API token from environment variables or secure command-line overrides rather than hardcoding sensitive information.\n- [SAFE]: The shell script utilizes proper quoting for all variables and restricts the API base URL to a known-safe default, mitigating risks of shell injection or unauthorized redirection.
Audit Metadata