Skill UI

This guide outlines the comprehensive reference architecture for building advanced, AI-powered search and research applications, modeled after Perplexity. It covers three core tiers: a searchable service with model routing and caching, a citation pipeline for validation and rendering, and an orchestrator for multi-query research synthesis. Use this when designing complex, production-grade, search-augmented LLM integrations or reviewing advanced system patterns.

This sophisticated tool generates publication-quality AI illustrations and diagrams for academic papers. It implements a multi-stage, iterative workflow, using advanced LLMs (like Claude and Gemini) to parse user requests, optimize complex layouts, verify academic styling (e.g., CVPR/NeurIPS standards), and render high-fidelity figures. Ideal for researchers needing professional, complex visuals for their manuscripts.

This meta-skill manages the end-to-end generation of AI-powered short dramas from a given topic. It first drafts a detailed, shot-by-shot shooting script, pausing for user review and adjustments. After confirmation, it generates consistent visual assets (universal character identity and scene compositions) and renders the final video clips. The output includes the fully rendered MP4, subtitles, title cards, and the complete script file.

Generates or edits single images using advanced AI models via OpenRouter. Accepts descriptive text prompts for text-to-image generation (e.g., concept art, product renders) or an input image for image-to-image editing. Supports various aspect ratios and resolutions, making it ideal for professional visual content creation workflows.

Inspect and audit GEPA genomes via the @metaharness/darwin/gepa library. Load, validate, render system prompts, and analyze failure modes in run transcripts. Ideal for adopting evolved policies, CI gates, and debugging evolution runs.

A comprehensive guide for authors submitting papers on 3D reconstruction, geometry, depth, and neural rendering to top-tier conferences like 3DV. It helps authors strategically position their work, refine the narrative, structure evidence, and measure fit against specific academic community expectations.

This meta-skill automates the entire short-drama production pipeline, starting from a simple topic. The workflow analyzes the request to infer render style, character identity, and shot count. It first drafts a detailed, shot-by-shot shooting script and pauses for user review. After approval, it generates a universal cast reference image, per-shot composition images, and corresponding video clips. Crucially, it maintains character and scene consistency by anchoring every video to both the reference image and the composition image. Finally, it adds title/ending cards, burns subtitles, and de...

Automated visual tuning: a vision or video model rates rendered variants in a loop. Render labeled variants into one artifact, ask the model to rate them and suggest better values, render suggestions, pick the best, repeat until good. Ideal for aesthetics, timing, layout, color grading, and other subjective visual parameters without numeric metrics.

This skill fetches public webpages and converts them into clean Markdown format suitable for AI agents. It allows users to read, summarize, quote, or cite webpage content while preserving the original source URL. The workflow involves sending a GET request to a specific endpoint to retrieve the rendered content. It is designed for read-only access and does not support private pages or content modification.