Total 50,523 skills, AI & Machine Learning has 8481 skills
Showing 12 of 8481 skills
Benchmark vLLM or OpenAI-compatible serving endpoints using vllm bench serve. Supports multiple datasets (random, sharegpt, sonnet, HF), backends (openai, openai-chat, vllm-pooling, embeddings), throughput/latency testing with request-rate control, and result saving. Use when benchmarking LLM serving performance, measuring TTFT/TPOT, or load testing inference APIs.
Find AI models on Replicate using search and curated collections.
Prompting techniques for AI image generation and editing models on Replicate. Use when writing prompts for image models or building image generation features.
Assess patent novelty and non-obviousness against prior art. Use when user says "专利查新", "patent novelty", "可专利性评估", "patentability check", or wants to evaluate if an invention is patentable.
Insert AI-generated illustrations into documents in-place. After reading the document, globally plan insertion points, generate all images in parallel, and insert them back into the original document asynchronously. Supports cover images, custom aspect ratios, and three styles. Use when: Users request to generate illustrations for documents/articles/notes. Also trigger when user mentions: illustrations, generate images, document images, add images to articles.
Import a migration bundle into Starchild. Downloads from relay, validates, and loads all components using native tools.
Install, initialize, verify, and troubleshoot RTK (Rust Token Killer) for AI coding agents. Use when you need to reduce shell-command token output, confirm that the correct `rtk` binary is installed, choose between Homebrew, install.sh, or Cargo installation, wire `rtk init` for Claude Code, Codex, Gemini CLI, Cursor, Copilot, Windsurf, Cline, or OpenCode, or use compact wrappers such as `rtk git status`, `rtk read`, `rtk grep`, `rtk test`, `rtk lint`, and `rtk gain`. Triggers on: rtk, rust token killer, token saver cli, rtk init, rtk gain, codex rtk, gemini rtk, opencode rtk, claude hook token reduction.
Shortcut alias for /superplan. Produce higher-quality code by breaking a feature into small, focused tasks the coding agent can nail one at a time. Works like an engineering team: feature → milestones → ~30-min tasks with specific files, acceptance criteria, and dependencies. Each task runs in a fresh context — narrow scope, full attention, one git commit per task.
Azure AI Vision integration. Manage data, records, and automate workflows. Use when the user wants to interact with Azure AI Vision data.
Pack, share, and load context using Epismo context packs. Trigger on: 'pack this', 'new pack', 'get <id>', 'read <alias>', 'load my context', 'what context do I have', 'restore session', 'save this context', 'share with my team', 'pack this up', 'hand this off', 'publish this guide', 'organize my packs', or any intent to persist or retrieve knowledge across tools or sessions.
Extract frames from video files using ffmpeg for AI/LLM analysis. Use when (1) the user asks to analyze, describe, or summarize a video file, (2) the user wants to extract frames or screenshots from a video, (3) the user provides a video file (.mp4, .mov, .avi, .mkv, .webm, etc.) and asks questions about its visual content, (4) the user wants to identify scenes, objects, or events in a video, (5) the user wants timestamps overlaid on extracted frames for temporal reference. Converts video into JPEG frames that can be attached to LLM prompts as images. Requires ffmpeg on PATH. Supports scene-change detection, model-aware optimization (Claude/OpenAI/Gemini), quality presets (efficient/balanced/detailed/ocr), grayscale and high-contrast OCR mode, and automatic FPS calculation via --max-frames.
End-to-end SGLang SOTA performance workflow. Use when a user names an LLM model and wants SGLang to match or beat the best observed vLLM and TensorRT-LLM serving performance by searching each framework's best deployment command, benchmarking them fairly, profiling SGLang if it is slower, identifying kernel/overlap/fusion bottlenecks, patching SGLang code, and revalidating with real model runs.