ima-knowledge-ai（我是知识 AI）

IMA Knowledge AI

Purpose: This skill provides strategic knowledge to help agents make better decisions when using IMA Studio's content creation APIs. It does NOT make API calls directly — instead, it guides you to use ima-voice-ai, ima-image-ai, ima-video-ai more effectively.

When to Use This Skill

**Read this skill BEFORE calling any ima-*-ai skill** if you need guidance on:

1. Workflow Design — How to break down complex user requests into atomic tasks
2. Model Selection — Which model to choose based on task requirements
3. Parameter Optimization — How to set parameters for quality, cost, or speed

Example scenarios:

• User: "帮我做个宣传视频" → Read workflow-design.md first
• User: "用最好的模型生成" → Read model-selection.md to pick the right one
• User: "生成16:9的图片" → Read parameter-guide.md for aspect ratio support

Knowledge Structure

All knowledge files live under the references/ directory. When reading from other skills, use paths like ima-knowledge-ai/references/workflow-design.md or ~/.openclaw/skills/ima-knowledge-ai/references/<filename>.md.

This skill contains 12 knowledge modules (8 standalone + 4 modular directories):

Core Knowledge (1-8)

1. workflow-design.md

When to read: Complex user requests that need task decomposition

• Task decomposition strategies
• Dependency identification (e.g., script → voiceover → video)
• Multi-step workflow templates
• Common creation patterns

2. model-selection.md

When to read: Choosing between multiple models for a task

• Model capability matrix (image/video/voice)
• Cost vs. quality trade-offs
• Use case recommendations (budget/balanced/premium)
• Model limitations and workarounds

3. parameter-guide.md

When to read: Optimizing parameters for a specific task

• Resolution/aspect ratio guidelines
• Quality vs. speed trade-offs
• Common mistakes and fixes
• Parameter compatibility matrix

4. visual-consistency.md ⭐ NEW

When to read: Any image/video generation task involving series, characters, or scenes

• Why AI generation lacks visual consistency by default
• Identifying implicit consistency requirements
• Reference image workflow (Image-to-Image / Video-to-Video)
• Multi-shot coherence strategies
• Common mistakes and best practices

5. video-modes.md ⭐⭐ CRITICAL

When to read: ANY video generation task (MANDATORY before calling ima-video-ai)

• image_to_video vs reference_image_to_video (DIFFERENT concepts!)
• image_to_video = first frame to video (input becomes frame 1)
• reference_image_to_video = reference appearance to video (can change scene)
• Traditional two-step vs modern one-step workflow
• Fallback strategy when primary method fails
• Common mistakes (旺财案例)

6. long-video-production.md 🎬 ESSENTIAL FOR LONG VIDEOS

When to read: User requests video longer than 15 seconds (30s ad, 1min short, 3min promo)

• Why models are limited to 10-15 seconds
• Multi-shot capability (2-4 camera angles in one generation) 🆕
• Three-step workflow: Script → Generate shots → Edit/Stitch
• Visual asset preparation (characters, scenes, props)
• Shot-by-shot generation strategy
• Video editing and stitching techniques
• Complete case study: 1-minute fantasy short film

7. character-design.md 🎨 CHARACTER DESIGN / IP DEVELOPMENT

When to read: User needs character design, IP development, game/animation assets, turnaround sheets

• Character Design industry overview (games, animation, manga, IP)
• Reference-driven workflow (Master Reference → Variants)
• Turnaround sheets (front/side/back/3-4 views)
• Expression library (happy/angry/sad/surprised...)
• Outfit variants (casual/armor/costumes)
• Props & weapons reference images
• Action poses (idle/walk/run/attack...)
• Complete case study: RPG game character "Aria"

8. vi-design.md 🏢 VI DESIGN / BRAND IDENTITY

When to read: User needs VI design, brand identity, logo applications, visual guidelines

• VI (Visual Identity) system overview
• Foundation system (Logo / Color / Typography / Auxiliary graphics)
• Application system (Office / Store / Packaging / Advertising / Digital / Uniform)
• Reference-driven workflow (Foundation → Applications)
• Logo consistency requirements (highest level)
• Color palette management (Primary / Secondary / Neutral / Functional)
• Complete case study: "Morning Light Coffee" cafe VI (20+ deliverables)

9. best-practices/ ⭐⭐⭐ COMMERCIAL TEMPLATES (On-Demand)

When to read: Commercial advertising or artistic photography tasks

Structure: Index + 4 scenario files (load only what you need)

• README.md — Index with keyword matching (2 KB)
• jewelry.md — Jewelry & accessories commercial ads (3 KB)
• skincare.md — Skincare & cosmetics commercial ads (3 KB)
• perfume.md — Perfume & fragrance commercial ads (3 KB)
• cinematic-art.md — Cinematic vintage art photography (4 KB)

Usage: Read index first → Load only relevant scenario file

Token savings: 60-85% compared to loading all scenarios

10. color-theory/ 🎨 COLOR THEORY & CULTURAL SENSITIVITY

When to read: Any design task requiring color selection (logos, posters, brands, products)

Structure: Index + 7 modular files (load only what you need)

• README.md — Index with quick navigation (5 KB)
• color-psychology.md — 11 major colors' psychology & applications (12 KB) ⭐
• color-combinations.md — 5 pairing principles (2 KB)
• industry-guide.md — 10 industries' color preferences (1 KB)
• cultural-differences.md — 5 regions' basic differences (1 KB)
• global-regions.md — 4 regions' detailed guide (4 KB)
• religious-systems.md — 5 major religions' color symbolism (5 KB)
• application-strategy.md — IMA Studio color decision process (2 KB)

Usage: Read index → Load relevant modules based on task (target region/industry/religion)

Token savings: 70-90% compared to loading all 32 KB

11. design-pitfalls/ 🚫 DESIGN MISTAKES TO AVOID

When to read: Quality assurance for generated content (before final delivery)

Structure: Index + 4 scenario files (29 common mistakes by scene)

• README.md — Index with 5 core principles (3 KB)
• logo-design.md — Logo design pitfalls (10 rules) (7 KB)
• poster-banner.md — Poster/Banner pitfalls (8 rules) (5 KB)
• product-ecommerce.md — Product/E-commerce pitfalls (6 rules) (3 KB)
• web-ui.md — Web/UI pitfalls (5 rules) (3 KB)

Usage: Read index → Load scenario-specific pitfalls

Token savings: 60-80% compared to loading all 21 KB

12. color-trends-2026/ 📅 2026 COLOR TRENDS

When to read: Design tasks for 2026 market (stay current with trends)

Structure: Index + 5 time/region files (load by current month + target region)

• README.md — Index with time-based navigation (6 KB)
• annual-colors.md — Pantone Cloud Dancer / WGSN Teal / China Horse Red (4 KB)
• spring-summer.md — Mar-Aug trends: Cobalt Blue, Violet, Bright Pink (2 KB)
• fall-winter.md — Sep-Feb trends: Dark Luxury theme (1 KB)
• regional-differences.md — China/US/Southeast Asia differences (2 KB)
• industry-applications.md — Tech/Fashion/Food/Beauty/Home (1 KB)

Usage: Read index → Load current season + target region

Token savings: 75-90% compared to loading all 16 KB

Auto-loading strategy: Check current month → Load relevant season file automatically

Usage Pattern

User Request
  ↓
[ima-knowledge-ai] Query relevant knowledge
  ↓
Make informed decision
  ↓
[ima-*-ai] Execute API call with optimized parameters
  ↓
Success!

Example flow:

User: "帮我生成一张16:9的产品海报，要高质量"

Step 1: Read ima-knowledge-ai → parameter-guide.md
        → Learn: SeeDream 4.5 supports 16:9, Nano Banana Pro native support
        
Step 2: Read ima-knowledge-ai → model-selection.md
        → Choose: Nano Banana Pro 4K (best quality, 18pts)
        
Step 3: Call ima-image-ai with:
        --model-id gemini-3-pro-image
        --extra-params '{"aspect_ratio": "16:9", "size": "4K"}'
        
Step 4: Success! 🎉

Important Notes

1. **This skill does NOT replace ima-*-ai skills**

Use it as a consultant before executing tasks

1. Knowledge is based on production IMA Studio API (2026-02-27)

Models and parameters may change; always verify with list-models

1. Cost transparency

All recommendations include credit cost for user decision-making

1. No scripts in this skill

Pure knowledge — implementation is handled by other ima-*-ai skills

Quick Reference