The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core functionality relies on executing external Git and LLM command-line tools (Codex, Claude, Amp, Pi) to perform reviews. The provided test suite extensively uses subprocess calls to manage test repositories and mock engine interactions. This behavior is necessary for the skill's primary purpose as a development tool.\n- [CREDENTIALS_UNSAFE]: The skill contains test fixtures (e.g., tests/fixtures/typescript-sensitive-literals.ts) with synthetic credentials like API keys and tokens. These are explicitly labeled as 'FAKE' and are used solely to verify the review engine's ability to identify sensitive data in code patches.\n- [DYNAMIC_EXECUTION]: The testing framework uses Python's SourceFileLoader to dynamically load the autoreview script for unit testing. This is a standard and safe implementation for testing standalone scripts in a Python environment.\n- [SAFE]: The skill demonstrates a high level of security awareness by implementing robust reviewer isolation (including macOS sandboxing support), sanitizing environment variables to prevent leakage (e.g., excluding HOME and TMPDIR in certain contexts), and strictly enforcing that all output and configuration files remain outside the repository being reviewed.

autoreview