zuiqiang-shifu-depth-lab

Builds evidence-backed depth drafts and converts them into spoken-word scripts for knowledge content.

11|1|Updated Aug 3, 2026
One-click install
npx skills add https://github.com/2282199198/zuiqiang-shifu --skill zuiqiang-shifu-depth-lab-2282199198
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: zuiqiang-shifu-depth-lab
Source: https://github.com/2282199198/zuiqiang-shifu/tree/main/depth-lab
Command: npx skills add https://github.com/2282199198/zuiqiang-shifu --skill zuiqiang-shifu-depth-lab-2282199198

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Creators of knowledge-based short videos often produce scripts that stay at the level of common sense, lack causal explanation, cite no verifiable sources, or sound like generic AI output. This Skill turns a confirmed topic into a researched depth draft with mechanism chains, evidence ledgers, and competing explanations, then converts it into a natural spoken script. ## Core Features & Use Cases - Depth Draft Construction: Builds an eight-part internal draft covering surface question, content type (decision/mechanism/research), non-obvious judgment, competing explanations, a three-plus-link causal chain, an evidence ledger, counterexamples, and audience takeaways. - Evidence Discipline: Separates verified facts, source opinions, inferences, and unknowns; supports source-restricted, web-research, and hypothesis-draft modes; refuses to fabricate data, cases, or statistics. - User-Controlled Expression Gate: Recommends 2-4 angles, duration (75-90 seconds or 2-3 minutes), and whether to fuse viral-style expression, then waits for the user's choice before writing the final script. - Use Case: A creator has chosen the topic "why premium hotels block food delivery riders" and wants a deeper script. The Skill proposes angles, builds a mechanism chain around business-model mismatch with sourced evidence, and delivers a readable 90-second spoken script after the user confirms the style. ## Quick Start Tell the AI to read SKILL.md and WORKFLOW_PROTOCOL.md, then provide your confirmed knowledge topic, target audience, and desired duration so it can recommend angles and wait for your choice before drafting.

Frequently Asked Questions about zuiqiang-shifu-depth-lab

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

FAQPage Schema
How do I write a deeper knowledge-based video script?

Provide a confirmed topic and let the Skill build a depth draft first: a non-obvious judgment, a three-plus-link causal chain, competing explanations, and an evidence ledger. Only after you confirm the angle, duration, and expression style does it write the final spoken script.

How does the Skill handle facts and sources in scripts?

It classifies every claim as verified fact, source opinion, inference, or unknown, and records each key fact in an evidence ledger with source, date, and scope. It supports source-restricted, web-research, and hypothesis-draft modes, and never fabricates data, cases, or statistics.

When should I not use this depth module?

Do not use it for title polishing, light rewrites, social posts, emotional or comedy content, or when you only need a quick direct answer. It also does not handle topic discovery; if you have no confirmed topic yet, use the topic-engine module instead.

Can I control the script style and length?

Yes. Before drafting, the Skill recommends 2-4 angles, a duration of 75-90 seconds or 2-3 minutes, and whether to fuse viral-style expression, then waits for your decision. Saying "use your recommendation" confirms all three; it never picks a style without your authorization.

What happens when evidence is insufficient for my topic?

The Skill narrows the conclusion instead of fabricating support. It delivers minimal verified facts, candidate hypotheses, evidence gaps, a small low-risk experiment, and stop conditions, and records second-order effects as checked-but-not-applicable with the specific gap noted.