deep-document-processor

Installation
SKILL.md

Deep Document Processor

Purpose

Transform large documents into token-efficient context artifacts optimized for downstream agent reasoning. The goal is NOT human-readable summaries — it's maximum decision-relevant signal per token.

Why This Exists

Agents face a fundamental tension: large documents contain critical context, but including them wholesale wastes tokens and dilutes attention. Naive approaches fail in predictable ways:

  • Full inclusion: Blows token budget, buries signal in noise
  • Naive truncation: Loses structure, misses late-document insights
  • Single-pass summary: Misses cross-references, flattens hierarchy

This skill applies a disciplined multi-pass protocol inspired by expert reading heuristics, adapted for agent context engineering.

Related skills

More from andurilcode/craftwork

Installs
4
GitHub Stars
6
First Seen
Apr 2, 2026