The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess module to execute local Git commands and LLM-provider CLI tools such as Codex and Claude. This execution is fundamental to its role as a code review adapter and is managed using sanitized environment variables to prevent local configuration leakage.- [INDIRECT_PROMPT_INJECTION]: Since the skill processes untrusted code changes from Git repositories, it has an inherent attack surface for indirect prompt injection. The skill includes documentation and system-level instructions aimed at the LLM to ensure reviewed code is treated as data rather than instructions.- [DYNAMIC_EXECUTION]: The unit testing framework (e.g., scripts/autoreview_test.py and tests/test_output_readiness.py) uses dynamic modules and code compilation to verify the main helper script's behavior in isolation. This use of dynamic loading is limited to the test suite and local development tasks.

openclaw-autoreview