harness-distillator

What problem does it solve?

This Skill solves the problem of wasting tokens and losing critical context when feeding many source documents into an LLM workflow, by producing hyper-compressed, lossless distillates that preserve every fact, decision, constraint, and relationship.

Core Features & Use Cases

• Lossless document compression: Converts source documents into token-efficient distillates optimized for LLM consumption (not lossy summaries).
• Semantic splitting with token budgets: Produces a single dense distillate or a semantically split multi-part distillate when size constraints demand it.
• Completeness verification and optional round-trip losslessness: Ensures extracted headings/entities appear in the output, and (optionally) reconstructs the originals to detect gaps or hallucinations.
• Purpose-aware filtering: Uses downstream_consumer context to drop irrelevant items while never dropping decisions, rejected alternatives, constraints, open questions, or scope boundaries.

Quick Start

Use the harness-distillator skill to distill all files matching 'docs/*.md' for a downstream workflow that creates PRDs.