codex-configuration

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION] (MEDIUM): The skill documents configurations that allow an agent to bypass security controls, specifically 'sandbox_mode = "danger-full-access"' and 'shell_environment_policy.inherit=all'. These settings can expose sensitive host environment variables and grant unrestricted disk access to the agent.
  • [REMOTE_CODE_EXECUTION] (MEDIUM): The instructions for configuring MCP servers promote the use of 'npx -y' to fetch and run remote packages. This enables arbitrary code execution at runtime based on the package name provided in the configuration.
  • [EXTERNAL_DOWNLOADS] (LOW): The skill references a built-in '$skill-installer' tool for downloading skills from GitHub. While standard for this tool, it highlights a vector for introducing unvetted external code into the runtime environment.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 03:59 AM