imagegen

Pass

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill reads OAuth access and refresh tokens from a local configuration file (~/.codex/auth.json). These credentials are sent to official OpenAI domains (chatgpt.com and auth.openai.com) to authenticate image generation requests. This is a functional requirement for the skill's operation and is clearly documented.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-defined prompts and processes local image files as style or subject references. This provides an attack surface where instructions embedded in external files could attempt to influence the behavior of the backend image generation process.
  • [COMMAND_EXECUTION]: Users can configure the backend API endpoint and orchestration model via environment variables or command-line flags. While providing legitimate flexibility for developer use, these parameters could be exploited if an agent is tricked into redirecting sensitive authentication data to an untrusted external server.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 16, 2026, 10:28 PM