The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands such as git diff and codex review to perform its primary function of code analysis. These operations are restricted to the local repository context.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to install the @openai/codex package. Since this package is within the @openai scope belonging to a trusted organization, it is considered a safe dependency reference.\n- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted data from git diff as part of the AI review process. \n
Ingestion points: Data enters the agent context via git diff output.\n
Boundary markers: No explicit delimiters are used to wrap the diff content in the shell command.\n
Capability inventory: The skill can execute local shell commands and write a review file to the repository.\n
Sanitization: No filtering or sanitization of the input diff is performed. This is documented as an inherent architectural trait of a code review tool and does not escalate the severity.\n- [CREDENTIALS_UNSAFE]: The skill mentions the OPENAI_API_KEY environment variable as a configuration requirement. It does not attempt to hardcode, log, or exfiltrate this credential.

codex-review