Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Generates complete BAML source files for type-safe LLM extraction, classification, RAG, and agent workflows from natural language requirements. Queries BoundaryML repositories via MCP for up-to-date patterns. Outputs typed client code for Python, TypeScript, Ruby, and Go.
BAML acts as a transpiler rather than a runtime dependency, so generated code has no BAML overhead in production and works natively with existing frameworks.
When to use it
- Extracting structured invoice data from unstructured text
- Building a classification pipeline for support ticket routing
- Creating a RAG function that returns answers with source citations
- Wiring a multi-step LangGraph agent to typed LLM outputs
- Generating a vision extraction function from image inputs
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: generate a BAML extraction function for parsing invoices from PDFs, using GPT-4o, with FastAPI integration
- 1Query baml_Examples MCP server for extraction patterns matching invoice and fastapi markers
- 2Define Invoice class with typed fields, @description and @assert annotations
- 3Generate Extract function with string and image? inputs targeting the Invoice class
- 4Configure a GPT-4o client with exponential retry policy
- 5Output complete .baml files plus a FastAPI endpoint and pytest test suite
baml_src/invoice.baml with Invoice class, Extract function, GPT4o client, retry policy, plus app/extract_invoice.py FastAPI route and tests/test_invoice.py with 100% function coverage
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.