After you've profiled your code and found the slow parts, this skill researches what good performance actually looks like for your domain, then digs into industry solutions for your specific bottleneck type (batching for architecture issues, connection pooling for I/O, algorithm alternatives for CPU). It checks your codebase first to see if you already have unused batch APIs or cache infrastructure before recommending new dependencies. The anti-bias checks are smart: it forces you to consider optimizing a slow feature before removing it entirely, which matters when "industry doesn't do X" conflicts with what your users actually need. Outputs 3-7 ranked hypotheses with expected impact, complexity, and file targets for the executor to try.
npx -y skills add levnikolaevich/claude-code-skills --skill ln-812-optimization-researcher --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills