The Agent Skills Directory

[COMMAND_EXECUTION]: The autoreview script supports a --parallel-tests feature that executes arbitrary shell command strings using subprocess.Popen(shell=True). This allows developers to run test suites concurrently with AI reviews but grants the agent direct shell access.
[COMMAND_EXECUTION]: The helper scripts invoke multiple system binaries including git, gh, and various review engine CLIs (e.g., codex, claude, cursor-agent). The skill includes custom logic to resolve these executables from the system PATH while explicitly ignoring any binaries located within the repository currently being reviewed to mitigate path-hijacking attacks.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest untrusted data in the form of code diffs and file content from a git repository. This creates a vulnerability where a malicious patch could contain instructions designed to manipulate the agent's evaluation or influence the arguments passed to the review tool (such as the test command).
[DATA_EXPOSURE]: The skill aggregates repository changes into a 'Change Bundle' and transmits this data to external AI providers (such as OpenAI or Anthropic) for analysis. While this is the intended purpose, it involves sending potentially sensitive local code to third-party services.

auto-review