The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the codex CLI tool using the Bash tool. The implementation uses safe command construction by piping content via stdin (using cat and printf) rather than passing data as command-line arguments, which mitigates the risk of argument injection.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the Codex CLI from OpenAI's official GitHub repository (https://github.com/openai/codex). This is a reference to a well-known and trusted organization's public resource.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data such as repository diffs, plan files (.claude/plans/), and design specifications (specs/spec.md).
Ingestion points: Reads local files and previous assistant responses to be reviewed by the external CLI.
Boundary markers: Uses a rubric file to provide specific instructions to the Codex tool. The skill explicitly instructs the agent to critically evaluate the tool's output and decide which points to adopt, reject, or defer.
Capability inventory: The skill utilizes Read, Write, and Bash tools to manage temporary files and execute the review process.
Sanitization: Data is piped to the external executable via stdin to prevent shell injection or manipulation of the command structure.

codex-review