llm-prompting

What problem does it solve?

This Skill provides expert guidance on LLM prompting infrastructure, patterns, and conventions used across the project's data-pipeline and evaluation components, enabling consistent design and safer, more reliable outputs.

Core Features & Use Cases

• Versioned Prompt System: frontmatter-based prompts with version tracking, hashing, and load/validate utilities.
• LLM Client Architecture: multi-provider routing and task-based model selection with deterministic fallbacks.
• Schema Enforcement: JSON validation via Pydantic models to guarantee structured outputs.
• Category Calibration: domain-specific benchmarks injected into prompts to improve calibration.
• Robust Quality Gating: evaluation and judge modules to ensure output reliability and governance.

Quick Start

Follow this guide to implement versioned prompts, category calibration, and schema enforcement in your LLM tooling.