gpt-image
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the Codex CLI installer from official OpenAI domains (
https://chatgpt.com/codex/install.shandhttps://chatgpt.com/codex/install.ps1). These sources are official platform distributors and the download is part of the intended setup process. - [COMMAND_EXECUTION]: The skill executes local system commands to manage its environment, including
node,git, and thecodexCLI for image generation tasks. It also executes the downloaded installer script during the bootstrap process to set up the Codex bridge. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided image prompts and file references. To mitigate potential injection risks, it implements a 'Prompt Fidelity Contract' that wraps user input in explicit boundary markers (
BEGIN USER PROMPT) when passing data to the Codex bridge. It also enforces path safety checks (isWithin) to prevent directory traversal attacks.
Audit Metadata