bmad-distillator

Compress multi-file source documents into lossless, token-efficient distillates for LLM workflows.

Updated Apr 2, 2026
One-click install
npx skills add https://github.com/amitmishra-byldd/Bmad-Note-Maker --skill bmad-distillator-amitmishra-byldd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bmad-distillator
Source: https://github.com/amitmishra-byldd/Bmad-Note-Maker/tree/main/_bmad/core/bmad-distillator
Command: npx skills add https://github.com/amitmishra-byldd/Bmad-Note-Maker --skill bmad-distillator-amitmishra-byldd

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Converts messy, multi-file source documents into a single lossless, token-efficient distillate that preserves every fact, decision, constraint, and relationship while removing human-oriented prose and noise so downstream LLM workflows can consume the result as sole context.

Core Features & Use Cases

  • Lossless extraction and compression: preserves specific numbers, dates, decisions, rejected alternatives, constraints, and relationships while removing rhetorical and filler text.
  • Analyzer-driven routing and semantic splitting: runs an analysis pass to recommend single vs fan-out compression, groups related files, and produces semantically coherent sections when needed.
  • Verification and validation: completeness checks, format checks, and optional round-trip reconstruction to prove fidelity; returns structured JSON results for integrations.
  • Use case: distill a product brief plus discovery notes into a compact, machine-loadable distillate for PRD creation, architecture design, or as input to an agent pipeline.

Quick Start

Generate a lossless distillate from the folder containing your source documents and save the result adjacent to the primary source with validation enabled.

Frequently Asked Questions about bmad-distillator

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

FAQPage Schema
How do I compress multiple documents into a single context for LLM processing?

Document distillation compresses multi-file source collections into a single, token-efficient context for LLM consumption. It removes human-oriented prose and noise while preserving every fact, decision, and constraint for downstream workflows.

What is semantic splitting and when do I need it for document distillation?

Semantic splitting divides distilled documents into semantically coherent parts. An analyzer-driven routing pass evaluates your source files and recommends either single dense compression or fan-out semantic splitting when downstream LLM workflows require structured context.

Can I validate that my compressed documents retain all original information?

Yes, lossless distillation supports completeness checks and optional round-trip reconstruction to prove fidelity. Validation ensures the compressed distillate preserves specific numbers, dates, rejected alternatives, and relationships from the original source documents.

How do I deduplicate overlapping content across multi-file research reports?

Deduplication during distillation identifies and removes overlapping content across multi-file document collections. The analyzer groups related files and eliminates redundant information while maintaining every unique fact and constraint for LLM consumption.

Does document distillation work with product briefs and discovery notes for PRD creation?

Distillation works with multi-file collections like product briefs, discovery notes, research reports, and architecture docs. It produces a compact, machine-loadable distillate optimized for creating PRDs, designing architecture, or feeding agent pipelines.

What's the best way to prepare source documents for lossless LLM compression?

Source documents require no special preparation before lossless LLM compression. The distillation process performs thematic analysis, compression, deduplication, and semantic splitting directly on messy, multi-file collections to produce a dense, machine-loadable context.