codex-pet
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions provide shell scripts that perform directory management, file manipulation, and automated image processing using the RunComfy and ImageMagick command-line tools.
- [EXTERNAL_DOWNLOADS]: The workflow requires downloading a reference image from an external HTTPS URL provided by the user to serve as the subject for the generated asset.
- [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by processing untrusted image data from external URLs through an AI image-editing model.
- Ingestion points: The skill ingest images from a user-provided SOURCE_URL during the generation process.
- Boundary markers: No delimiters or instructions are used to separate the image content from the model's instructions.
- Capability inventory: The skill has the ability to execute shell commands, perform network operations via a CLI, and write files to the local file system.
- Sanitization: The skill does not perform validation or content filtering on the input image before it is processed by the external API.
- [PRIVILEGE_ESCALATION]: Prerequisite steps in the documentation suggest using system package managers like apt-get for installation, which typically require administrative privileges.
Audit Metadata