codex
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill generates shell commands for the
codexCLI and utilizes stdin piping (e.g.,echo "prompt" | codex) which introduces a potential shell injection surface if the agent does not properly escape user-provided content. - [COMMAND_EXECUTION]: The skill supports high-privilege configuration options such as the
--sandbox danger-full-accessflag, enabling broad system and network access; these are mitigated by strict instructions to seek user consent viaAskUserQuestionbefore use. - [PROMPT_INJECTION]: The instructions require the agent to append
2>/dev/nullto commands by default to suppress thinking tokens, which may hide execution errors or security warnings from the user's view. - [SAFE]: The skill follows operational safety standards by instructing the agent to stop on non-zero exit codes and verify follow-up actions with the user after every command execution.
Audit Metadata