If you're working with LLMs and need guaranteed valid output formats, this is worth checking out. It enforces structure at the grammar level during generation, so you get valid JSON, XML, or code every time without post-processing or retries. Works with Pydantic models for type safety and supports local models through Transformers, llama.cpp, and vLLM. The speed claim is real because it constrains token sampling directly rather than generating then validating. Originally from dottxt.ai with 8,000+ GitHub stars. Most useful when you're building agents or tools that can't afford malformed responses breaking your pipeline.
npx -y skills add davila7/claude-code-templates --skill outlines --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