bmad-distillator

Extract lossless, token-efficient distillates from source documents.

Updated Mar 25, 2026
One-click install
npx skills add https://github.com/JingyiWng/databricks_ai_dev_kit_price_watcher --skill bmad-distillator-jingyiwng
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bmad-distillator
Source: https://github.com/JingyiWng/databricks_ai_dev_kit_price_watcher/tree/main/_bmad/core/skills/bmad-distillator
Command: npx skills add https://github.com/JingyiWng/databricks_ai_dev_kit_price_watcher --skill bmad-distillator-jingyiwng

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Distillator enables lossless compression of source documents into a single distillate (or semantically partitioned distillates) that downstream LLM workflows can consume without losing information, addressing the need to minimize token usage while preserving facts, decisions, and constraints.

Core Features & Use Cases

  • Lossless distillates: preserve all source information with maximal token efficiency.
  • Multi-stage workflow: analyze, compress, verify, and optionally round-trip validate distillates.
  • Flexible outputs: single distillate beside sources or a folder of sectioned distillates.

Quick Start

Provide one or more source_documents and optional flags to generate a distillate next to the primary source.

Frequently Asked Questions about bmad-distillator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I compress LLM context tokens without losing source document information?

Lossless distillation compresses source documents into token-efficient distillates while preserving all facts, decisions, and constraints. This minimizes token usage for downstream LLM workflows without sacrificing information, utilizing a multi-stage workflow to analyze, compress, and verify outputs.

What is the best way to verify that distilled LLM context retains all original facts?

Round-trip validation verifies distilled LLM context by triggering end-to-end checks using the --validate option. This enforces frontmatter-driven metadata and required inputs to ensure the compressed distillate remains completely lossless compared to the original source documents.

Can I generate multiple sectioned distillates instead of a single compressed file?

Yes, flexible outputs allow generating either a single distillate beside the primary source or a folder of semantically partitioned, sectioned distillates. This organizes dense LLM context into manageable partitions while maintaining maximal token efficiency.

Does the distillation workflow support frontmatter-driven metadata for LLM inputs?

Yes, the distillation workflow enforces frontmatter-driven metadata and required inputs for LLM contexts. This ensures structured tracking of the lossless compression stages, from initial document analysis and compression to final verification and optional round-trip validation.

What are the limitations of using lossless distillation for dense LLM context?

Lossless distillation for dense LLM context requires providing one or more source documents to function. While it maximizes token efficiency, the multi-stage analysis, compression, and verification process may introduce processing overhead compared to simple truncation or basic summarization methods.