Skill UI

Reference for building Graph Neural Networks with PyTorch Geometric, covering data objects, datasets, transforms, convolution layers, lazy initialization, and heterogeneous support so you can quickly tackle node, link, and graph tasks.

A comprehensive skill for leveraging massive pre-trained transformer models across diverse AI domains, including Natural Language Processing (NLP), Computer Vision (CV), and Audio. Use this skill for advanced tasks such as text generation, classification, question answering, summarization, object detection, and model fine-tuning on proprietary datasets. It provides streamlined pipelines and granular control for deployment.

This tool is designed for defining and scaffolding production-grade, multi-agent workflows. It allows developers to move beyond single prompts by structuring complex tasks into explicit steps. Core capabilities include selecting workflow patterns (e.g., sequential, parallel, orchestrator), defining handoff contracts, implementing error recovery strategies, and enforcing context and cost discipline across long-running AI processes.

Evaluates incoming requests, decides if specialized skills are needed, orchestrates optimal combinations, and logs successful pairings in agent-memory while guarding against overusing complex tools on simple tasks.

Coordinate multiple AI agents concurrently via dmux’s tmux pane manager, running Claude Code, Codex, OpenCode and other tools in parallel tracks for research, implementation, testing, review, or multi-role coding tasks.

Detect malicious scheduled task creation and modification using Sysmon (Event IDs 1/11), Windows Security events (4698/4702), and correlation of suspicious parents, public paths, and encoded arguments for persistence and lateral movement analysis.

Run inspect-ai or lighteval evaluations on Hugging Face Hub models locally, guiding backend choice, smoke tests, and GPU fallbacks before handing off remote tasks.

Run Hugging Face models directly in JavaScript/TypeScript for text, vision, and speech tasks without backend servers, using Transformers.js in Node.js or modern browsers with WebGPU, quantization, and pipeline abstractions for efficient inference.

Provides command-line recipes for hitting DeerFlow HTTP endpoints to start threads, stream runs, manage memory/models/skills/agents, and upload files for in-depth research tasks.

Switches the AI persona into a Chinese-style mama nagging narrator that sticks to core behavior guardrails while delivering increasingly intense motivational scripts; useful when you need a relentless reminder that you must research, verify, and finish tasks before asking for help.

Automate Datadog monitoring tasks via Rube MCP’s Composio toolkit—query metrics, search logs, manage monitors/dashboards, and create events or downtimes while respecting live tool schemas.

Coordinates qwenpaw agent workflows by identifying which peer agent’s expertise or workspace context is needed, then initiating chat or background tasks with the prescribed session, task, and parameter rules.