SDO optimization: renamed displayName and rewrote description for search discoverability. Lexical boost 0 → +1.1 on all target queries.

Fix: set displayName via --name flag (was stuck on auto-generated slug-based name)

ima-knowledge-ai v1.0.5 - Skill name updated for clarity to: "Knowledge Base for AI Video, Image & Audio Generation" - No changes to knowledge modules or content structure - Documentation (SKILL.md) improved for clearer purpose and usage instructions - No impact on API, only metadata and descriptive enhancements

# Changelog — ima-knowledge-ai All notable changes to the **IMA Knowledge AI** skill will be documented in this file. --- ## [1.0.4] - 2026-03-11 ### Changed - Optimized Midjourney usage guide - Improved SKILL.md header metadata --- ## [1.0.3] - 2026-03-05 ### Changed - Optimized skill description - Model selection guide: Improved Midjourney and Nano Banana 2 logic --- ## [1.0.2] — 2026-03-04 ### 📚 Modular Knowledge Base Expansion **Major update: Added 68 KB of new knowledge with modular on-demand loading architecture.** #### Added **New Modular Knowledge Directories** (contributor: 李鹤, senior designer): 1. **`references/color-theory/`** (32 KB, 8 files) - Color psychology for 11 major colors - 5 color combination principles - Cultural differences (China/US/Japan/India/Middle East + 4 detailed regions) - Religious color systems (5 major religions) - Industry-specific color preferences (10 industries) - IMA Studio color decision strategy 2. **`references/design-pitfalls/`** (21 KB, 5 files) - 29 common design mistakes across 4 scenarios - Logo design pitfalls (10 rules) - Poster/Banner pitfalls (8 rules) - Product/E-commerce pitfalls (6 rules) - Web/UI pitfalls (5 rules) 3. **`references/color-trends-2026/`** (16 KB, 6 files) - 2026 annual representative colors (Pantone Cloud Dancer, WGSN Transformative Teal, China Horse Red) - Spring-Summer trends (Mar-Aug: Cobalt Blue, Violet, Bright Pink) - Fall-Winter trends (Sep-Feb: Dark Luxury theme) - Regional differences (China/US/Southeast Asia) - Industry-specific applications **Modular Architecture**: - Each directory has an index `README.md` with quick navigation - Agent loads only relevant modules (60-90% token savings) - Relative path linking between modules - Time-based and scenario-based loading strategies #### Changed - **SKILL.md**: Updated to reference 12 knowledge modules (was 8) - Added sections 10-12 for new modular directories - Expanded Quick Reference table (+7 entries, now 28 total) - Updated version: 1.0.1 → 1.0.2 - **Total knowledge base**: 116 KB → 184 KB (+68 KB, +59%) - **Total files**: 13 → 42 (+29 files) #### Fixed - **Security & Compliance**: - Removed brand name "ImaClaw" → generic "评价/建议" - Replaced operational instructions with descriptive guidance (ClawHub security scan compliance) - Removed ffmpeg code snippets → "use video editing tools" - Removed ima_prefs.json file reference → "consider user preferences" - Changed specific tool names → generic descriptions - **Documentation**: - All references remain strictly advisory (no operational steps) - Maintained focus on planning and parameter choices #### Technical Details **Contributor Credit**: 李鹤 (Senior Designer) - 2.5 hours professional knowledge contribution (2026-03-03) - 90 KB professional design knowledge - Color theory + cultural sensitivity + design pitfalls **Commit History**: - `9837d4f` - Initial 3 documents (64 KB) - `198986c` - Modularized color-theory + design-pitfalls - `a56c97b` - Modularized color-trends-2026 - `7301a66` - Updated SKILL.md accessibility - `242c67b` - Removed ImaClaw brand name - `2960cc1` - Security compliance fixes --- ## [1.0.1] — 2026-03-03 ### 🔧 ClawHub Release Fixes **Cleanup and improvements for public release.** #### Removed - **Development Scripts**: Removed `scripts/parse_arena_leaderboard.py` (development-only tool, not needed for production) #### Changed - **Best Practices Links**: Updated `references/best-practices/README.md` to use relative markdown links - Scene index table now uses `[jewelry.md](jewelry.md)` format instead of inline code - Keyword sections now link to respective scene files - Improves navigation in ClawHub skill browser #### Fixed - Version consistency across all files (1.0.0 → 1.0.1) --- ## [1.0.0] — 2026-03-03 ### 🎉 Initial ClawHub Release This is the first public release of **ima-knowledge-ai** — a comprehensive knowledge base for strategic guidance on IMA Studio multi-media content creation. ### 📚 Knowledge Base (9 Topics, ~80 KB optimized) #### Core Strategic Guidance 1. **workflow-design.md** (7.2 KB) - Task decomposition strategies - Dependency identification methods - Multi-step workflow templates - Common creation patterns 2. **model-selection.md** (8 KB) ⭐ **NEW** - Task type consolidation (7 types: 2 image, 4 video, 1 music) - Arena.AI leaderboard rankings (Text-to-Image, Text-to-Video) - Model recommendations with real performance data - Content policy warnings (OpenAI real-person restrictions) - Midjourney special notes (strong aesthetics, weak text rendering) 3. **parameter-guide.md** (8 KB) ⭐ **REWRITTEN** - Task type awareness (modification tasks vs new generation) - Prompt optimization rules (when to optimize, when not to) - Aspect ratio selection strategy - Midjourney special implementation (--ar parameter) #### Visual Consistency & Production 4. **visual-consistency.md** (12 KB) - Why AI lacks consistency by default - Reference-driven generation workflow - Image-to-Image / Video-to-Video modes - Multi-shot coherence strategies 5. **video-modes.md** (8 KB) ⭐ **OPTIMIZED** - `image_to_video` vs `reference_image_to_video` (critical distinction!) - Traditional two-step vs modern one-step workflow - Fallback strategies when primary method fails 6. **long-video-production.md** (8 KB) ⭐ **OPTIMIZED** - Why models are limited to 10-15 seconds - Shot-by-shot generation strategy - Video editing and stitching techniques - Simplified workflow patterns 7. **character-design.md** (8 KB) ⭐ **OPTIMIZED** - Character Design workflow (Reference-driven) - Turnaround sheets, expression library, outfit variants - Action poses and consistency strategies 8. **vi-design.md** (8 KB) ⭐ **OPTIMIZED** - VI (Visual Identity) system overview - Foundation system (Logo / Color / Typography) - Application system (Office / Store / Packaging / Digital) - Reference-driven workflow (Foundation → Applications) #### Best Practices (Modular Structure) ⭐⭐⭐ **NEW** 9. **best-practices/** (15 KB, 5 files) - **On-demand loading structure** for context efficiency - **Index + 4 scenario files** (jewelry, skincare, perfume, cinematic-art) - Contributed by 李鹤 (colleague) - **Token savings: 60-85%** per task (load only relevant scenario) Files: - `README.md` (2 KB) — Index with keyword matching - `jewelry.md` (3 KB) — Jewelry & accessories commercial ads - `skincare.md` (3 KB) — Skincare & cosmetics commercial ads - `perfume.md` (3 KB) — Perfume & fragrance commercial ads - `cinematic-art.md` (4 KB) — Cinematic vintage art photography --- ## 🎯 Core Methodology ### Reference-Driven Generation **The universal principle** taught across all advanced topics: > Generate a **Master Reference** first → Use it to generate all **Variants** This methodology applies to: - **Video Production** → Master character/scene → All shots - **Character Design** → Base design → Turnaround sheets, expressions, outfits - **VI Design** → Logo foundation → All application materials - **Commercial Ads** → Master visual style → Product variations **Why it works**: - AI models generate **random variations** by default - Reference images provide **visual anchors** for consistency - `reference_strength` parameter controls consistency level (0.7-0.95) --- ## 📊 Knowledge Base Optimization ### Before Optimization (Initial Version) - **Total Size**: 184 KB (8 files) - **Longest file**: 34 KB (long-video-production.md) - **Structure**: Single monolithic documents ### After Optimization (v1.0.0) - **Total Size**: ~80 KB (9 topics, -53% reduction) - **Longest file**: 12 KB (visual-consistency.md) - **Structure**: Modular best-practices (on-demand loading) - **Token efficiency**: 60-85% savings per task ### Optimization Highlights - ✅ 4 files slimmed: 172 KB → 80 KB (-53%) - ✅ parameter-guide.md rewritten: 502 lines → 208 lines (-59%) - ✅ model-selection.md rewritten: Arena.AI leaderboard data integrated - ✅ best-practices split: Single 13.3KB file → 5 modular files (2-4KB each) **Philosophy**: Agent documentation = Decision handbook, not textbook --- ## 🌟 Key Features ### What This Skill Does ✅ **Strategic Guidance** — Helps you make better decisions ✅ **Model Selection** — Recommends optimal models for tasks ✅ **Parameter Optimization** — Teaches cost/quality/speed trade-offs ✅ **Visual Consistency** — Reference-driven workflow methodology ✅ **Production Workflows** — Real-world case studies with step-by-step guides ✅ **Cost Transparency** — Clear credit costs for all recommendations ### What This Skill Does NOT Do ❌ **API Calls** — This is pure knowledge, not execution ❌ **File Operations** — No image/video generation or uploads ❌ **Direct Content Creation** — Use `ima-image-ai`, `ima-video-ai`, `ima-voice-ai` for that --- ## 🔗 Related Skills **ima-knowledge-ai** works alongside IMA Studio execution skills: - **[ima-image-ai](https://git.joyme.sg/imagent/skills/ima-image-ai)** — Image generation (text-to-image, image-to-image) - **[ima-video-ai](https://git.joyme.sg/imagent/skills/ima-video-ai)** — Video generation (text-to-video, image-to-video) - **[ima-voice-ai](https://git.joyme.sg/imagent/skills/ima-voice-ai)** — Music generation (text-to-music) - **[ima-all-ai](https://git.joyme.sg/imagent/skills/ima-all-ai)** — Unified multi-media generation - **[ima-resource-upload](https://git.joyme.sg/imagent/skills/ima-resource-skill)** — File upload to IMA OSS --- ## 🎓 Target Audience ### Who Should Use This Skill? - ✅ **AI Agents** using ima-*-ai skills - ✅ **Content Creators** planning multi-step workflows - ✅ **Designers** working on character/IP development - ✅ **Brand Managers** creating VI/identity systems - ✅ **Video Producers** making long-form content - ✅ **Developers** integrating IMA Studio APIs - ✅ **Anyone** needing strategic guidance for AI content creation --- ## 🛠️ Technical Details ### Knowledge Base Structure ``` references/ ├── workflow-design.md # 7.2 KB — Task decomposition ├── model-selection.md # 9.7 KB — Model recommendations ├── parameter-guide.md # 12 KB — Parameter optimization ├── visual-consistency.md # 12 KB — Reference-driven workflow ├── video-modes.md # 31 KB — Video generation modes ├── long-video-production.md # 34 KB — Long-form video guide ├── character-design.md # 22 KB — Character/IP design └── vi-design.md # 31 KB — VI/brand identity ``` ### API Version Compatibility - **Based on**: IMA Studio Production API (2026-02-27) - **Models Covered**: 20+ models across image/video/music - **Last Verified**: 2026-03-02 --- ## 🚀 Installation ### Via ClawHub CLI ```bash clawhub install ima-knowledge-ai ``` ### Manual Installation ```bash cd ~/.openclaw/skills git clone https://git.joyme.sg/imagent/skills/ima-knowledge-ai.git ``` --- ## 📖 Usage Pattern ``` User Request ↓ [ima-knowledge-ai] Query relevant knowledge ↓ Make informed decision (model, parameters, workflow) ↓ [ima-*-ai] Execute API call with optimized settings ↓ Success! 🎉 ``` --- ## 💡 Example Scenarios ### Scenario 1: Image Series **User**: "生成一套产品图,5张不同角度" **Knowledge consulted**: - `visual-consistency.md` → Reference-driven workflow - `parameter-guide.md` → Optimal resolution settings **Result**: Generate 1 master reference → Use it for 4 additional angles --- ### Scenario 2: Long Video **User**: "做个1分钟的宣传片" **Knowledge consulted**: - `long-video-production.md` → Multi-shot workflow - `video-modes.md` → Shot generation modes - `visual-consistency.md` → Maintain visual coherence **Result**: Script → 6 shots (10s each) → Video editing → 1min output --- ### Scenario 3: Character Design **User**: "游戏角色设计,需要多角度视图" **Knowledge consulted**: - `character-design.md` → Turnaround sheet workflow - `visual-consistency.md` → Reference-driven generation **Result**: Master design → 3-4 view turnaround sheet → Expression library --- ## 🎯 Future Roadmap ### Planned Topics (v2.0+) - **Prompt Engineering** — Advanced prompt writing techniques - **Cost Optimization** — Budget control and batch generation strategies - **Failure Handling** — Error recovery and retry strategies - **Case Studies Library** — More real-world production examples - **Performance Optimization** — Speed vs. quality trade-offs --- ## 📞 Support & Feedback - **Issues**: [GitLab Issues](https://git.joyme.sg/imagent/skills/ima-knowledge-ai/-/issues) - **Discussions**: [ClawHub Comments](https://clawhub.com) - **API Support**: [IMA Studio](https://imastudio.com) --- ## 📜 License MIT License — See [LICENSE](LICENSE) for full details. --- ## 🙏 Credits **Developed by**: IMA Skills Team **Contributors**: OpenClaw Community **Special Thanks**: All early testers and feedback providers --- **Last Updated**: 2026-03-03 **Version**: 1.0.1 **Status**: ✅ Ready for ClawHub Release --- **"Knowledge is power — but only when applied!"** 🍵

Optimized skill description; Model selection guide: Improved Midjourney and Nano Banana 2 logic

- Documentation refreshed and reorganized for improved clarity in SKILL.md and reference files. - Minor corrections and formatting updates throughout guides and best-practices. - Updated best-practices/README.md for more accurate scenario keyword mapping. - Removed deprecated script: scripts/parse_arena_leaderboard.py. - No changes to the strategic guidance content or usage patterns.

ima-knowledge-ai 1.0.0 — Strategic IMA Studio guidance for workflow, model, and parameter decisions - Introduces a comprehensive knowledge base for IMA Studio content creation, covering workflow design, model selection, and parameter optimization. - Adds 8+ modular reference files, including new critical guides for visual consistency, video generation modes, long video strategies, and commercial templates. - Provides best practices for decomposing complex tasks and choosing optimal parameters/models for image, video, and voice API workflows. - Ensures users avoid common mistakes and understand trade-offs across quality, cost, speed, and model limitations. - Designed as a consultant layer—guides decision-making before executing any ima-*-ai skill.