The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates for executing the codex command-line tool to analyze source code changes. These commands are standard for the intended purpose of code review.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code content, which is a standard surface for indirect prompt injection in analysis tools.
Ingestion points: Source code accessed via --uncommitted, --base, and --commit flags in SKILL.md.
Boundary markers: Absent; instructions do not specify delimiters to separate code from commands.
Capability inventory: Shell execution of the codex command in SKILL.md.
Sanitization: Not specified in the instructions.
[SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. The functionality is consistent with the stated purpose of the skill.

codex-review