The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the @openai/codex CLI tool, which is an external dependency downloaded from the npm registry. This is a legitimate dependency on a well-known service required for the skill's functionality.
[COMMAND_EXECUTION]: The skill executes a bash script (run.sh) that runs several git commands to compute differences between branches and identify untracked files.
[COMMAND_EXECUTION]: The installation instructions for the required codex CLI suggest using sudo for a global install. This practice allows the installation script to run with elevated root privileges, which is a privilege escalation risk.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it includes untracked filenames directly in the model's prompt.
Ingestion points: Untracked filenames retrieved via git ls-files in run.sh.
Boundary markers: Absent; filenames are listed as plain text within the prompt.
Capability inventory: The codex exec tool allows the language model to read file contents and run commands in a read-only environment.
Sanitization: None applied to filenames before inclusion in the prompt.

local-review-codex