The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @openai/codex package via NPM (npm install -g @openai/codex). This is a legitimate request for a CLI tool from a well-known service provider, and the installation is initiated by the user rather than executed silently by the script.
[COMMAND_EXECUTION]: The skill uses standard development commands such as git, gh (GitHub CLI), and mktemp for repository state management and output capture. These operations are restricted to the local environment and are consistent with the skill's stated purpose of reviewing code changes.
[INDIRECT_PROMPT_INJECTION]: The skill processes user questions and file contents through the codex exec command. It implements boundary markers by explicitly defining the tool's persona (e.g., "brutally honest technical reviewer") and instructing it to be terse. The use of the -s read-only flag further mitigates risk by ensuring the external model cannot modify the local file system during analysis.

codex-consult