The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell scripts (audit-start.sh, audit-next.sh) to automate environment setup. These scripts execute standard system commands including git worktree for repository isolation, mkdir for directory management, and the codex CLI tool to launch agent sessions with specific task-oriented prompts. These operations are restricted to the local filesystem and the vendor's own tools.
[PROMPT_INJECTION]: The skill instructions in templates/GOAL.md define a robust 'Audit-only' persona with strict operational constraints. While the skill is designed to process untrusted code from repositories (creating a surface for indirect prompt injection), it mitigates this risk by explicitly forbidding code modifications during the audit phase and restricting the agent's write capabilities to a single tracking file (PROGRESS.md).
[SAFE]: No external network operations, obfuscation, persistence mechanisms, or credential harvesting were detected. The skill's behavior is entirely local, transparently implemented in scripts, and consistent with its stated purpose of facilitating secure code reviews.

code-quality-audit