Learn: AI Short Drama Script Creation (微短剧剧本)

Reverse-engineering a highly structured, 7-step Claude Code workflow for creating monetizable, industry-standard short drama scripts.
01 MAR 2026 • ERIC SAN • DONNA ATTEMPT R1

I. TL;DR + Verdict

A WeChat article revealed a highly structured AI "Skill" for Claude Code that writes 50-100 episode short drama scripts (微短剧). Instead of standard "write a story" prompts, it forces the AI through a 7-step pipeline (Concept -> Outline -> Characters -> Index -> Episodes -> QA -> Export) bounded by rigorous pacing and "dopamine" rules.

DONNA MASTERS

What it means: The artifact is essentially an AI agent prompt architecture. By extracting the core industry theory (Golden 5-seconds, Paywalls, 30-60-30 pacing) and encoding it into a Cursor Skill, Donna can now execute this exact 7-step process on command.

Leverage: Extremely high. It bridges the gap between our current video production capabilities (`vidprod`) and narrative creation, specifically tuned for modern attention spans and monetization formats.

II. The Artifact

SOURCE: WeChat Article: "短剧剧本创作 Skill 正式发布..."
CREATOR: A professional short drama team utilizing Claude Code.
WHAT IT IS: A packaged AI workflow (Skill) that generates production-ready short drama scripts, including a "Go-Global" mode for international platforms like ReelShort.

What makes it special? (The Gap)

We already have robust video production prompts (`vidprod`), but they focus heavily on cinematography, storyboard parsing, and visual execution. This artifact focuses purely on hyper-optimized narrative pacing and psychological hooks (爽点/卡点). It does not treat the script as literature; it treats it as a precision-engineered dopamine delivery system with hard constraints (e.g., paywalls at Ep 10-15, cliffhangers every 60 seconds).

III. How It's Done (The Deep Research)

The workflow succeeds because it embeds 1,500+ lines of industry "dry goods" (干货) into the AI's context window. Key mechanisms:

IV. Donna's Reproduction Attempt

ATTEMPTED: Reverse-engineering the 7-step workflow and knowledge base into a native Cursor Skill.
RESULT: Succeeded.

What Worked

Blockers & Gaps

None that prevent execution. The skill is live. The only fundamental limit is Taste (Human Judgment)—while the AI can follow the structure, Eric must evaluate if the "Hook" actually feels compelling to a human audience during the /自检 (Self-QA) step.

V. Prerequisite Map

Capability Status Gap Remaining To Close
Script Mechanics & Formatting Mastered None. The skill enforces format strictly. Use .cursor/skills/short-drama-script
Pacing Theory (卡点/爽点) Mastered None. AI context is loaded with theory. AI automatically applies during generation.
Taste Assessment Partial AI cannot truly "feel" if a scene is emotionally satisfying. Eric's Role: Review during Steps 2 (Outline) and 5 (Episode Draft).

VI. Mastery Path + Testing Next Steps

Donna has cleared the road. The capability is now native to the workspace.

Eric's Quick Test (< 30 min)

  1. Open a new chat with Donna.
  2. Type: /开始 战神归来题材,出海模式 (Start God of War genre, Go-Global mode).
  3. Let Donna guide you through Step 2 (Outline) and Step 3 (Characters).
  4. Evaluate the generated Paywall placement and Western trope adaptation.

VII. Critical Assessment

Is this just a prompt wrapper? Yes and no. While it is "just" a prompt architecture, the value is in the constraints. By forcing a 7-step pipeline rather than a single generation, it prevents the AI from losing the plot or writing "AI-flavored" generic prose.
The "Already Know It" Trap: We have vidprod and know how to prompt for scripts. However, this artifact proved we were missing the brutal, hyper-optimized commercial structure of the Micro-Drama format (the 30-60-30 pacing, the exact placement of paywalls). By remaining infinitely curious, we acquired a highly specific monetization framework, not just "scriptwriting."