codex
Warn
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run the
codexCLI with high-impact flags, including--sandbox danger-full-accessand--full-auto. These flags grant the external tool broad system access and autonomous editing capabilities. - [COMMAND_EXECUTION]: Instructions in
SKILL.mdmandate appending2>/dev/nullto shell commands by default. This suppresses the standard error stream, which prevents the user from seeing security warnings, tool errors, or indicators of unexpected behavior during execution. - [COMMAND_EXECUTION]: The skill recommends a shell piping pattern (
echo "prompt" | codex ...) for processing user input. This construction is vulnerable to command injection if the agent does not rigorously escape special shell characters within the user-provided prompt. - [COMMAND_EXECUTION]: The skill explicitly instructs the agent to always use the
--skip-git-repo-checkflag, which bypasses the tool's built-in safety checks regarding the state and location of the repository being analyzed.
Audit Metadata