openai-codex-security

Warn

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Installs the @openai/codex-security package from the NPM registry. This package is attributed to a trusted organization.
  • [COMMAND_EXECUTION]: Utilizes CLI tools and shell commands like npx and npm to perform security scans and manage local authentication state, including integration with system keyrings for credential storage.
  • [REMOTE_CODE_EXECUTION]: Describes an 'Automated Fix Application' workflow where the agent interprets results from the security scan and applies code changes directly to source files. Applying non-deterministic AI-generated modifications to a codebase is a form of dynamic execution that carries significant operational risk.
  • [PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection by processing external data from the local codebase.
  • Ingestion points: All files in the targeted scan directory (e.g., .) processed via the codex-security scan command or SDK run method.
  • Boundary markers: The instructions lack explicit delimiters or guidance for the model to ignore embedded commands within the scanned content.
  • Capability inventory: Full directory read access and write access for code fixes.
  • Sanitization: No defined process for sanitizing or escaping code content before it is processed by the underlying AI model.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 30, 2026, 12:38 AM
Security Audit — agent-trust-hub — openai-codex-security