krystal-draft

Generate paired verbatim and editorial draft artifacts preserving collaboration provenance.

Updated Apr 11, 2026
One-click install
npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill krystal-draft-gozonerd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: krystal-draft
Source: https://github.com/gozonerd/shadow-ai-assessment/tree/main/.claude/skills/krystal-draft
Command: npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill krystal-draft-gozonerd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When arguments are developed through AI-human dialogue, it becomes impossible to prove which words are the human's own and which came from the AI. This Skill solves that provenance problem by producing two paired artifacts: a verbatim human draft and a separate editorial companion, so authorship is structural rather than remembered. ## Core Features & Use Cases - KR (Krystal Draft): Extracts the user's verbatim responses from a designated conversation corpus, preserving typos exactly, with zero word-level edits, and assigns stable human-readable IDs (MLN-topic-slug-NN) to each idea. - CC (Claude Companion): Captures all editorial support the user did not explicitly engage — hardenings, sharpenings, honest-gap bridges, and confirm-for-KR items — cross-referenced to the same shared ID space. - Strict provenance rules: Refuses to edit verbatim spans, bridge missing argumentative steps inside KR, emit unanchored editor bullets, or publish one artifact without the other. - Use Case: After several chat sessions refining an argument for an essay or position paper, invoke the skill to produce <artifact>_KR_<date>_v01.md and <artifact>_CC_<date>_v01.md with a provenance front-matter statement on both. ## Quick Start Say "krystal draft" and designate the artifact target and source conversation span to generate the paired KR and CC draft files.

Frequently Asked Questions about krystal-draft

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

FAQPage Schema
How do I create a draft that preserves my exact words from a chat conversation?

Invoke the skill with a trigger like "krystal draft" and explicitly designate the source conversation span. It extracts your verbatim responses with typos intact, selects and orders them into the KR artifact, and never performs word-level edits.

What is the difference between the KR and CC artifacts?

KR contains only the user's verbatim words, selected and ordered but never edited. CC contains all of the AI's editorial suggestions the user did not explicitly engage, including hardenings, sharpenings, and honest-gap bridges, cross-referenced to the same idea IDs.

Can the skill fix typos or add connective text in the verbatim draft?

No. The skill explicitly refuses to fix typos, add [sic] markers, or bridge missing argumentative steps inside KR. Any connective or clarifying content goes into the anchored bullet layer or the CC companion instead.

What happens if I don't designate a source conversation?

The skill halts and asks for the source corpus rather than inferring it. It never harvests text from ambient or unmarked conversation spans, and research or AI-generated content is categorically excluded from the verbatim layer.

When should I not use this drafting approach?

Skip it for deliverables with no argument spine, such as logs, indices, or mechanical outputs. It is also the wrong tool when you want polished prose, since that requires a separate deliverable with different provenance handling.