The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands (git diff, git status, git show) to retrieve code changes for review. These operations are restricted to the local repository and are essential for the skill's primary function.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it processes untrusted source code through the git diff and git show commands. 1. Ingestion points: git diff and git show output in SKILL.md Phase 1. 2. Boundary markers: Absent. 3. Capability inventory: Uses Agent tool for parallel review tasks and writes reports to the local .claude/ directory. 4. Sanitization: Absent. However, the skill implements strong mitigations by using read-only subagents with restricted instructions and requiring a human-in-the-loop to verify and manually execute any resulting implementation plans.
[SAFE]: No data exfiltration, hardcoded credentials, or persistence mechanisms were detected. All file operations are restricted to local workspace reports and plans.

ac-review