The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the execution of external binaries such as Git, TruffleHog, and CLI interfaces for LLM engines including Codex, Claude Code, and Pi.
[COMMAND_EXECUTION]: The --parallel-tests feature enables the execution of user-defined test commands within a temporary, isolated workspace.
[CREDENTIALS_UNSAFE]: Test fixtures in the tests/fixtures/ directory contain hardcoded mock API keys and tokens designed to validate the efficacy of the integrated secret scanner.
[DATA_EXFILTRATION]: The skill includes functionality to stage credentials for remote delegation during testing when the OPENCLAW_TESTBOX environment variable is set; this is an intended feature for trusted maintainers.
[PROMPT_INJECTION]: The instructions provide a scope governor and contract that ensure the agent treats review output as advisory and manually verifies findings to mitigate risks from indirect prompt injection in analyzed code.

autoreview