manju-laoli-skill
Industrial-grade multi-agent screenwriting & audiovisual directing system for viral short/animated dramas, with 6-stage Asset-First pipeline adapted to major AI agent platforms
Suggested prompts
About this skill
Manju Laoli AIGC Full-Pipeline Skill
A workflow for screenwriting, asset management, audiovisual direction, generation prompts, and quality control for short dramas and animated short dramas. It follows an Asset-First order: lock character, scene, and prop assets before storyboarding and prompt rendering.
Production Pipeline
Project lock-in → gated screenwriting → dialogue diagnosis → asset locking → dialogue/action storyboards → prompt rendering → independent quality control
Core Capabilities
| Stage | Capability |
|---|---|
| Pipeline | A P0–P5 six-stage production order in which asset images come before storyboards |
| Screenwriting | A five-stage Premise → Structure → Beat → Entity → Page gate that prevents dialogue from being written before structure |
| Dialogue | Seven-dimension diagnosis, three speech-speed tiers, and a shot-splitting workflow intended to reduce broadcast-style delivery |
| Assets | A/B/C graded locking; CHR, AUD, PRP, SCN, and Uxx ledgers; T1→T2 family-lineage derivation |
| Emotion | A 12-beat emotional-tension curve with visual plotting |
| Storyboards | Separate methods for micro-expression progression in dialogue scenes and action direction, including R1/R2/R3 scoring for 15-second action completion |
| Action | Rules covering 23 martial-arts schools, nine sword forms, weapons, lightness skills, and chained finishing moves |
| Fantasy | Rules for artifacts, five-element magic, spell attack and defense, visible energy, and army-scale action |
| Rendering | Seedance 2.5-oriented rules for 15-second clips, first/last-frame dual images, Hit-Stop, [SFX:] audio, and three-layer prompt separation |
| Space | A 2×2 top-view camera map plus templates for dialogue blocking and combat positions |
| Quality control | P0/P1/P2 gates and a six-step P4 pre-feed checklist |
| Compliance | A safety-language dictionary intended to reduce platform risk, without guaranteeing approval |
Workflow
- Define the project, aspect ratio, and production constraints; explicit user instructions take priority for horizontal or vertical output.
- Build the premise, structure, beats, entities, and page-level script.
- Diagnose dialogue, speech speed, and shot splitting.
- Create and lock ledgers for characters, voices, props, scenes, and other assets.
- Enter the dialogue-scene or action-storyboard path and establish camera and character-position maps.
- Generate layered prompts for Seedance.
- Pass independent quality gates before feeding prompts into production.
What You Provide
Provide the genre and narrative goal, target platform, horizontal or vertical format, duration, and any existing script, characters, scenes, props, or references. If assets already exist, identify which elements must remain locked.
What You Get
- A gated story structure and script
- Dialogue diagnosis and speech-speed checks
- Character, voice, prop, and scene asset ledgers
- An emotional-tension curve
- Dialogue or action storyboards and camera maps
- Generation prompts for Seedance 2.5
- Pre-feed quality-control results
Requirements and Limitations
- The Skill can be called by AI assistants or LLMs that support Agent Skills. The source README explicitly lists adapters for OpenClaw, WorkBuddy, Doubao Coze, and Dify.
- The package includes Python scripts for integrity checks and emotion-curve plotting.
- The safety dictionary is intended to reduce risk and does not guarantee platform approval.
- Actual generation quality depends on the selected model, reference assets, and the quality of the provided information.