codex-images
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
@openai/codexpackage via the npm registry or through Homebrew. These are official distribution channels for a well-known service provider, following the established trust scope for external dependencies. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text to formulate prompts for the Codex image generation tool. It lacks explicit sanitization or boundary markers for these inputs; however, this is a standard functional surface for an image-generation agent, and the instructions specifically caution the agent to check the resulting text for accuracy.
- [COMMAND_EXECUTION]: The skill utilizes several standard command-line tools, including
mkdir,cp,ls,shasum,sips,ffprobe, andffmpeg. These tools are used for project file organization, data integrity verification, and image transformations (scaling and cropping) within the local workspace. - [SAFE]: The skill demonstrates good security posture by explicitly instructing the agent never to ask for OpenAI keys or passwords in the chat, relying instead on the user's local authentication via the
codex logincommand.
Audit Metadata