App Store Screens

Guide for using Microsoft MarkItDown - a Python utility for converting files to Markdown. Use when converting PDF, Word, PowerPoint, Excel, images, audio, HTML, CSV, JSON, XML, ZIP, YouTube URLs, EPubs, Jupyter notebooks, RSS feeds, or Wikipedia pages to Markdown format. Also use for document processing pipelines, LLM preprocessing, or text extraction tasks.

Translates an image (or a set of image references — screenshots, mockups, Figma URLs, live websites) into two mirrored design-system artifacts: `docs/design.md` (YAML tokens + prose, following Google's open [design.md](https://github.com/google-labs-code/design.md) format, for the coding agent) and `docs/design.html` (a self-contained, token-driven style guide rendering every token and component live, for the human to read). Reads the imagery, asks targeted clarifying questions, derives the design tokens (colors, typography, spacing, rounded, components), and writes both files. Fully standalone — requires no other document or skill. Use when the founder says "create a design system", "design from image", "translate image to design", "create design.md", "image to design system", "extract design tokens", or shares an image with no other clear intent.

Use when the user asks for kiss cam or a task matching the examples below. Generate a viral fake "in-arena Kiss Cam moment" of any two subjects — a fan-filmed phone shot of the MSG Jumbotron with retro Kiss Cam graphic + scoreboard, plus a 15s Kling v3-omni clip with PA-announcer commentary and crowd reaction. Any subject styles (human, 3D toy, illustrated avatar). No names. Triggers: "make me a kiss cam moment", "kiss cam version of these two", "Jumbotron kiss cam trend", "fake NBA kiss cam". Requires the pika MCP.

Fine-tune any HuggingFace CV / VLM / LLM model on local NVIDIA GPUs inside an NGC PyTorch container. Use when the user wants to fine-tune a HuggingFace model (full or LoRA), train a vision / VLM / LLM model end-to-end, generate a reproducible HF training pipeline, smoke-test a HuggingFace model locally before scale-up, push a fine-tuned model to the HF Hub with a model card, or emit a self-contained rerun skill for an existing HuggingFace finetune. Supports image classification, object detection, semantic / instance / panoptic segmentation, depth estimation, image-text-to-text VLM (SFT / LoRA), and LLM SFT / DPO / GRPO. Six-step workflow: inspect and qualify, hardware and NGC image, research, generate and smoke, train + eval + infer, push and emit rerun skill.

Reusable template for authoring new Agent Skills with clear triggers, workflow, and I/O contracts.

Cosmos-Embed1 video-text embedding for text-to-video retrieval, video-to-video search, semantic deduplication, and fine-tuning. Use when the user asks to "fine-tune Cosmos-Embed1", "run cosmos-embed inference", "export Cosmos-Embed1", "embed videos", or "search videos with text".