codex

Warn

Audited by Gen Agent Trust Hub on Apr 23, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is centered around executing the codex-wrapper command-line utility to perform automated code tasks.
  • [EXTERNAL_DOWNLOADS]: The documentation specifies that the tool should be obtained from GitHub Releases or an install.sh script, which represents a dependency on an unverified third-party source.
  • [PROMPT_INJECTION]: The skill's feature for multi-file analysis (using @ syntax) creates a surface for Indirect Prompt Injection. Malicious content within analyzed files could attempt to override the agent's instructions.
  • Ingestion points: File content from the local environment and user-defined task text.
  • Boundary markers: Uses shell HEREDOC to prevent local shell injection, but lacks delimiters to isolate external file content within the AI prompt.
  • Capability inventory: File access and execution of the wrapper binary.
  • Sanitization: There is no mention of sanitizing or validating the contents of the files before they are processed by the AI.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 23, 2026, 10:55 AM