meitu-image-fix

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: Orchestrates complex image repair pipelines by executing the vendor's meitu-cli toolset (including image-upscale, image-edit, image-cutout, and image-beauty-enhance) through shell commands.
  • [EXTERNAL_DOWNLOADS]: Utilizes curl as a fallback mechanism to download processed image assets from remote URLs returned by the tool's API when direct URL passing fails.
  • [COMMAND_EXECUTION]: Invokes a local workspace management script (oc-workspace.mjs) to resolve output paths, manage project metadata, and verify environment configurations.
  • [PROMPT_INJECTION]: Ingests user-provided text instructions that are interpolated into tool command arguments (e.g., the --prompt flag in image-edit), which serves as a potential surface for indirect prompt injection if inputs are not sanitized by the platform.
  • [CREDENTIALS_UNSAFE]: References the use of environment variables (OPENAPI_ACCESS_KEY, OPENAPI_SECRET_KEY) and CLI configuration commands for managing API credentials, which is standard practice for authenticating with vendor services.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 11:58 PM