codex-image-gen
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The legacy Python implementation in scripts/generate.py uses subprocess.run to call the system curl utility. This is used for reliable API communication and includes logic to restrict access to temporary files containing bearer tokens.
- [EXTERNAL_DOWNLOADS]: The skill performs network operations to official OpenAI and ChatGPT domains (openai.com and chatgpt.com) to generate images and refresh access tokens.
- [CREDENTIALS_UNSAFE]: The tool programmatically accesses authentication tokens and configuration from ~/.codex/auth.json and ~/.codex/config.toml to use the user's existing sessions.
- [PROMPT_INJECTION]: The skill processes user-provided strings for image generation, forming an indirect prompt injection surface. * Ingestion points: User-supplied text via the --prompt argument in scripts/generate.mjs and scripts/generate.py. * Boundary markers: None. * Capability inventory: Local file writes and network requests to OpenAI endpoints. * Sanitization: Relies on the remote provider's moderation and safety filters as documented in SKILL.md.
Audit Metadata