The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs git operations such as add -N and reset to toggle the visibility of untracked files. It correctly utilizes the -- delimiter in bash commands to protect against argument injection from potentially malicious file names.
[DATA_EXFILTRATION]: The skill is designed to expose the contents of untracked files to a chain of review tools. Users should ensure that sensitive files (e.g., .env or keys) are gitignored or excluded via the Scope parameter to prevent unintended exposure to the automated review process.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes untracked file data.
Ingestion points: Untracked files discovered via git status in SKILL.md.
Boundary markers: Absent; file content is interpreted by the secondary review skills.
Capability inventory: Bash for git manipulation and Skill dispatch for review orchestration.
Sanitization: Includes a post-execution validation step that re-parses YAML frontmatter to detect and warn about corruption caused by called skills.

review-with-untracked