crystallization

Distill completed work chains into durable digests and reusable lessons.

Updated Apr 20, 2026
One-click install
npx skills add https://github.com/thistleknot/skills --skill crystallization
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: crystallization
Source: https://github.com/thistleknot/skills/tree/main/crystallization
Command: npx skills add https://github.com/thistleknot/skills --skill crystallization

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Crystallization converts finished, purposeful work into a durable digest and reusable lessons so the same conclusions do not need to be rediscovered every session.

Core Features & Use Cases

  • Distill completed work into a durable digest: extract a clear question, findings, entities, lessons, and open questions from a completed work chain.
  • Extract and route reusable knowledge: send digest pages and semantic facts to agentic_kg_memory, and route reusable skill-contract deltas to skill-wiki.
  • Manage lesson lifecycle with confidence and deduping: save lessons with initial confidence (0.6) and reinforce via a compounding rule, using fingerprint deduplication to avoid duplicates.
  • Avoid crystallizing noise: gate on verified completeness and stable outcomes to prevent canonizing speculation.

Quick Start

Use crystallization to distill a completed research or debugging work chain into a question-driven digest and extracted lessons, then route the resulting artifacts to agentic_kg_memory and skill-wiki as appropriate.

Frequently Asked Questions about crystallization

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

FAQPage Schema
How do I turn completed research into reusable lessons and a digest page?▼

Distilling debugging lessons requires extracting a clear driving question, stable findings, and heuristic anchors from a completed work chain. Verification gating ensures the outcome is stable, preventing the canonization of speculative fixes into your knowledge base.

How does memory routing work for extracted research findings?▼

Deduplicating knowledge management entries uses fingerprint matching to identify existing lessons. New lessons are saved with an initial confidence score of 0.6 and reinforced via a compounding rule, ensuring repeated discoveries strengthen existing memory rather than creating duplicates.

When should I not use research distillation on a work chain?▼

You should avoid distillation when a work chain lacks verified completeness or stable outcomes. Applying synthesis to unstable or speculative results gates out noise, preventing premature conclusions from becoming durable lessons in your memory routing system.

What is the best way to distill multi-turn analyses into durable knowledge?▼

Distilling multi-turn analyses involves structured extraction of intent, discovery, ecosystem, and heuristic anchors. The process creates a question-driven digest page that outlives the current session, ensuring research convergence and named tradeoffs are preserved.