Building and training neural networks requires significant computational resources and distributed infrastructure. Flow Nexus provides cloud-based neural network training with support for multiple architectures including feedforward, LSTM, GAN, transformer, and autoencoder models across distributed E2B sandboxes.

Managing multiple AI agents working on the same codebase creates conflicts and coordination overhead. Agentic Jujutsu provides lock-free version control with self-learning capabilities that improve over time and resolve conflicts automatically.

Traditional databases struggle with semantic search queries. AgentDB delivers 150x faster vector operations with sub-millisecond search times for building intelligent RAG systems and knowledge bases.

AgentDB vector searches slow down as data grows, causing application lag. This skill provides quantization and indexing techniques to achieve 150x faster searches with 4-32x memory reduction while maintaining accuracy.

Building AI agents that forget context between conversations leads to poor user experiences. This skill provides AgentDB memory patterns for persistent storage, enabling agents to remember interactions, learn from experiences, and maintain context across sessions.

Building self-learning agents requires complex reinforcement learning setup. This skill provides templates and guidance for 9 algorithms including Decision Transformer, Q-Learning, and Actor-Critic.

Build sophisticated distributed AI systems with sub-millisecond synchronization and advanced search capabilities. Learn to implement QUIC sync, custom distance metrics, and hybrid vector plus metadata search for production applications.

Creating professional text-to-video content requires understanding model-specific parameters and capabilities. This skill provides guided access to Wan 2.7's audio-driven lip-sync and multi-reference conditioning through the RunComfy CLI.

Videos with wrong aspect ratios get rejected by platforms or look unprofessional. This skill uses AI to spatially extend video frames, converting vertical clips to horizontal or adding environment beyond original frames.

Removing objects, watermarks, or unwanted elements from video clips is difficult with standard editing tools. This skill routes your request to the right AI model on RunComfy for prompt-driven video inpainting across frames.

Short video clips limit creative storytelling and make it hard to build narrative sequences. This skill extends existing Veo clips with consistent motion, lighting, and subject identity using the RunComfy CLI and Google Veo 3-1 extend-video API.

Video editing requires matching the right AI model to your goal. This skill intelligently routes requests to Wan 2.7, Kling, or Lucy based on whether you need restyle, motion control, or identity-stable edits.

Creating professional short-form video with lip-sync and multi-modal references requires chaining multiple tools and managing complex input schemas. This skill wraps the RunComfy CLI to provide Claude with direct access to ByteDance Seedance 2.0 Pro, handling API calls, polling, and file downloads seamlessly.

Developers and creators need a single, scriptable interface for hundreds of AI image and video models. The RunComfy CLI provides one binary with one authentication to access every RunComfy model endpoint, from image generation to video editing and LoRA training.

Changing the lighting of a photograph traditionally requires reshooting or advanced Photoshop skills. This skill routes to purpose-built AI relighting models that change light direction, color temperature, and mood while preserving the subject's identity and composition.

Image editing workflows require precise identity preservation and batch processing capabilities. This skill provides expert guidance for using Google Nano Banana 2's edit endpoint to swap backgrounds, localize changes, and process up to 20 images per call.

Naive prompting against Nano Banana 2 wastes the model's strengths. This skill bundles the model's documented prompting patterns so Claude, Codex, and Claude Code produce sharper images on the first try — from rapid drafts to typography-perfect brand assets.

Lip-syncing faces to audio requires choosing the right model from many options. This skill routes requests across RunComfy's lip-sync endpoints automatically, selecting Sync Labs, OmniHuman, Kling, or Creatify based on your input type and quality needs.

AI agents need a reliable way to call advanced video generation models without writing custom API code. This skill gives Claude and Codex direct access to all six Kling 3.0 endpoints on RunComfy, from draft previews to 4K masters with native audio.

Creating videos from images requires selecting the right AI model for your use case. This skill intelligently routes your request to HappyHorse for portraits, Wan 2.7 for lip-sync, or Seedance for multi-modal animations.