The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the codex review command to analyze code modifications. It also suggests using standard Unix utilities like wc and tail for monitoring task output files (SKILL.md).
[PROMPT_INJECTION]: An indirect prompt injection surface exists because the skill processes untrusted code changes from git repositories.
Ingestion points: Git staged and committed changes read by the codex tool (SKILL.md).
Boundary markers: Absent; the instructions do not specify the use of delimiters to encapsulate untrusted data (SKILL.md).
Capability inventory: The skill possesses command execution capabilities (codex, wc, tail) which could be influenced by malicious code content (SKILL.md).
Sanitization: No sanitization or validation logic is defined for the input code before processing (SKILL.md).

codex-review