A wrapper around Anthropic's official SDK that adds structured outputs and schema validation, so Claude's responses match your exact format every time. The big win here is the outputFormat parameter that works with JSON schema or Zod, giving you full TypeScript inference and validated results via message.structured_output. Worth noting the breaking changes: v0.1.45 introduced structured outputs in November 2025, and you'll need the beta header structured-outputs-2025-11-13. The security audits are mixed (Gen Agent Trust Hub failed, Socket passed, Snyk warned), so review those before using in production. If you're tired of parsing messy LLM responses, this handles the validation layer for you.
npx skills add https://github.com/jezweb/claude-skills --skill claude-agent-sdk