gpt-image-2

Pass

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: References the OpenAI Codex CLI hosted at github.com/openai/codex as a prerequisite for functionality.
  • [COMMAND_EXECUTION]: Utilizes child_process.spawn to execute the codex command-line interface locally for image generation and editing tasks.
  • [DATA_EXFILTRATION]: Accesses the ~/.codex/sessions/ directory to retrieve image data from session rollout files. The skill implements a unique correlation marker and directory snapshotting to restrict data access to the specific files generated during the current request.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 29, 2026, 10:53 AM