prompt-engineering

To get structured JSON outputs from LLMs, you design prompt engineering workflows using schemas and Pydantic. This skill enables you to define JSON-based outputs, tool-call results, and system prompts to reliably elicit structured data from OpenAI, Anthropic, and open-source models.

The best way to write cross-provider prompts is by designing multi-provider portability workflows. This skill covers techniques to structure system prompts, personas, and tool use that function consistently across OpenAI, Anthropic, and open-source models like Llama.

Use zero-shot prompting for clear, deterministic instructions and few-shot prompting with examples when step-by-step reasoning is required. This skill provides workflows for both, alongside chain-of-thought techniques, to optimize large language model outputs for diverse tasks.

Yes, you can integrate RAG and function calling into your prompt engineering workflows. This skill enables tool use, multi-step workflow chaining, and RAG integration to orchestrate complex tool calls and generate reliable results from large language models.

You need LangChain and Pydantic to manage prompt versioning and structured outputs. This skill leverages these dependencies alongside OpenAI, Anthropic, tiktoken, and tenacity to orchestrate tool use and ensure reliable, high-quality outputs.