Cangjie Knowledge Distillation · Dad Edition

The Problem It Solves

Methodology-heavy works often contain reusable decision frameworks, principles, and checklists, but simply feeding them to an agent usually yields summaries rather than reliable trigger logic. This skill converts selected works into executable skills that preserve trigger conditions, evidence, and counter-examples, reducing vague summaries and hallucinated citations.

How It Works

• Whole-work understanding: reads files or URLs, outlines the structure, interpretation, critique, and application, then confirms the distillation scope.
• Parallel extraction: extracts frameworks, principles, examples, counter-examples, and terminology into candidate knowledge units.
• Quality filtering: applies cross-domain evidence, predictive power, and distinctiveness checks to remove content that only restates the source or states generic common sense.
• Construction and testing: generates SKILL.md definitions with trigger conditions, citation limits, and test-prompts.json covering should-trigger, should-not-trigger, and ambiguous edge cases.
• Publishing loop: after stress tests pass, syncs the skill to SkillHub and preserves audit trails and related-skill links.

Boundary Notes

It suits methodology, decision frameworks, thinker-based frameworks, and multi-source distillation. It is not intended for ordinary book reviews, roleplay, or simple translation. Input must come from accessible text; if the source is missing, it should stop and request the material rather than reconstruct it from memory.