gpt-image-2
Pass
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local bash and Python scripts to interface with the
codexCLI tool. It passes the user's prompt via standard input and handles file paths as script arguments, which is a standard method for integrating CLI tools without introducing shell injection vulnerabilities. - [DATA_EXPOSURE]: To retrieve generated images, the skill reads from the user's
~/.codex/sessions/directory. It mitigates potential data exposure by diffing the directory before and after execution, ensuring it only processes the specific session files generated during that session rather than accessing older chat history. - [EXTERNAL_DOWNLOADS]: The skill depends on the external
codexCLI tool andpython3being installed on the user's system. This dependency is documented as a prerequisite for the skill's core functionality. - [SAFE]: The skill includes a dedicated validation layer in
scripts/extract_image.pythat checks the proposed output path. It explicitly forbids writing to sensitive system directories (such as/etc,/usr, or/System) and restricts file output to authorized image extensions.
Audit Metadata