This runs a structured code review against diffs, not full codebases. It auto-detects whether you're working in frontend, backend, CLI, or config files based on directory paths, then loads the relevant checklists. The review covers security basics like hardcoded secrets and command injection, plus LLM-specific smells like placeholder implementations and overly generic abstractions. It also does impact analysis by searching for breaking changes to exports and APIs. Everything gets written to a timestamped markdown report in /tmp with severity levels and line numbers. Best used when you want consistent review criteria across PRs without manually remembering what to check for in each domain.
npx skills add https://github.com/llama-farm/llamafarm --skill code-review