codex-imagegen
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The README contains instructions for installing the Codex CLI by downloading and executing shell/PowerShell scripts directly from OpenAI's official domain (
chatgpt.com). This is a documented installation pattern for this service. - [COMMAND_EXECUTION]: The skill orchestrates image generation by spawning subprocesses for the
codexcommand and utility Python scripts (python3 -c) used for resizing and transparency conversion. The implementation uses shell quoting to sanitize parameters and prevent command injection. - [PROMPT_INJECTION]: The skill ingests user-defined visual prompts which are passed to the image generation backend. It implements boundary markers and parameter validation to manage the processing of this untrusted data.
- Ingestion points: User-provided
prompt,outpath, andiconslist. - Boundary markers: Prompts are delimited within the tool-generated instruction template using specific labels.
- Capability inventory: Subprocess execution for CLI tools and Python scripts; file system write access for output storage.
- Sanitization: CLI arguments are escaped using shell-quoting; output paths are resolved relative to the workspace.
Audit Metadata