@skills-shell

ai-avatar-video covers talking head video generation via inference.sh CLI across four model families. p-video-avatar leads with built-in tts, 18x speed advantage, and cost efficiency. includes workflows for dubbings, ugc batching, and multi-character scenarios.

youtube-thumbnail-design provides actionable guidance on dimensions, color contrast, safe zones, text placement, and face expression psychology for high-ctr thumbnails, with example belt cli prompts for ai generation across content types.

character-design-sheet provides techniques for maintaining visual consistency across ai-generated character images, covering turnaround views, expression sheets, LoRA training, and detailed description anchors. focuses on practical workflows via the belt CLI.

python-executor runs arbitrary python code in an isolated 8gb sandbox with 100+ preinstalled libraries (numpy, pandas, matplotlib, requests, pillow, trimesh, etc). outputs saved to outputs/ are auto-returned. timeout configurable 1-300s, high-memory variant available.

gpt-image wraps openai's gpt-image-2 model for text-to-image, editing, and inpainting via belt cli. covers multiple formats, batch generation, and custom resolutions but lacks error handling guidance and intent clarity.

infsh-cli wraps 250+ cloud-hosted ai models (image, video, llm, search, 3d, twitter) behind a single cli. installation is straightforward and examples cover common tasks, but the skill lacks actionable intent definitions and error handling guidance.

web-search provides tavily and exa integration for web queries and content extraction via belt cli, but lacks structured intent definition, explicit error handling, and trigger coverage beyond the quick start example.

agent-browser provides playwright-based web automation for ai agents with element reference (@e ref) interaction, but lacks structured intent definition, decision logic, and failure handling documentation.

agent-tools wraps a cloud inference CLI offering 250+ apps across image generation, video, LLMs, search, and twitter automation via simple command invocation, but lacks skill-level intent definitions and error handling guidance.

competitor-teardown provides a starting point for competitive analysis using belt CLI and tavily search, but lacks structured methodology, decision trees, or explicit output contracts. execution remains shallow and incomplete.

twitter-automation provides basic cli wrappers for twitter/x operations like posting and engagement, but lacks procedural depth, trigger clarity, and failure mode documentation.

product-hunt-launch provides a cli-based workflow for generating product hunt gallery assets via flux image generation, lacking strategic guidance on timing, tagline optimization, and launch day tactics promised in the description.

image-to-video provides a minimal cli-based workflow for converting still images to video via inference.sh, but lacks model selection guidance, motion prompting strategy, and deployment context.

p-image wraps pruna's optimized text-to-image models via inference.sh cli for straightforward prompt-based generation. setup is minimal but procedural detail and failure handling are sparse.

seedance wraps bytedance seedance 2.0 video generation via inference.sh cli, supporting text prompts with optional synchronized audio synthesis. minimal documentation and unclear failure handling limit production readiness.

product-photography wraps the inference.sh CLI to generate product images via a bytedance diffusion model. covers studio packshots, lighting, and e-commerce conventions, but lacks procedural depth and failure handling.

app-store-screenshots provides a shell interface for generating app store assets and preview videos, but the provided documentation is incomplete and lacks actionable procedures beyond tool installation.

p-video wraps pruna's optimized video generation models (p-video, wan-t2v, wan-i2v) through inference.sh cli with minimal documented guidance on input schemas, model selection, or failure recovery.

landing-page-design attempts to cover conversion optimization through ai image generation via a belt cli wrapper, but lacks coherent guidance on layout principles, cta mechanics, or design decision-making.

ai-image-generation wraps inference.sh CLI to invoke 50+ image models including flux and gpt-image-2, but lacks actionable procedure detail, intent specification, and failure handling guidance.

ai-video-generation wraps inference.sh CLI to invoke 40+ video models, but lacks procedural depth, decision logic, and failure handling. a thin passthrough without meaningful automation structure.

storyboard-creation wraps flux image generation via belt CLI to produce individual panels, but lacks structured guidance on shot vocabulary, continuity rules, panel sequencing, or filmmaking craft.

Generate and edit videos with Alibaba HappyHorse 1.0 models via inference.sh CLI. Models: HappyHorse T2V, I2V, R2V, Video Edit. Capabilities: text-to-video,…

Generate talking head avatar videos with Pruna P-Video-Avatar via inference.sh CLI. Turn a portrait image into a realistic speaking video with built-in TTS.…

Video ad creation with exact platform-specific specs for TikTok, Instagram, YouTube, Facebook, LinkedIn. Covers dimensions, duration limits, AIDA framework,…

Render videos from React/Remotion component code via inference.sh. Pass TSX code, get MP4. Supports all Remotion APIs: useCurrentFrame, useVideoConfig, spring,…