Skills

Operational guide for choosing and combining TP, DP, PP, CP, and EP parallelism strategies in Megatron Bridge, with sizing rules by model size and hardware topology for both dense and MoE models.

A skill from the claude-mem project, which provides persistent memory and session management capabilities for Claude Code with support for over 20 languages.

A skill from the bencium marketplace, a Claude Code plugin collection with 13 skills for design, architecture, and productivity. Covers systematic UX design, typography, accessibility (WCAG 2.1 AA), and engineering reviews.

Enables AI agents to perform crypto wallet operations including authentication, sending USDC/ETH/SOL, token trading, balance checking across Base/Polygon/Solana, and querying onchain data via the awal CLI.

A Claude Code skill ecosystem with 22 sub-skills for creating, optimizing, and managing blog content at scale, generating complete articles, editorial calendars, and structured data dual-optimized for Google rankings and AI citation platforms.

A product management skill that guides AI agents through finance-based pricing strategy analysis using battle-tested product management frameworks.

from phuryn/pm-skills

Guides selective and full activation recompute in Megatron Bridge to reduce GPU memory usage, with a prioritized checklist from expandable segments through core_attn, layernorm, and mlp recompute modules.

Converts HTML/CSS designs into fully formatted PowerPoint presentations, supports template-based slide creation, and enables precise visual editing of presentation files.

Converts single-node training scripts to multi-node Slurm sbatch jobs for Megatron Bridge, covering srun-native launches, Enroot container setup, NCCL timeout debugging, and MoE model OOM sizing.

Provides guidance for training MoE models at long sequence lengths in Megatron Bridge, covering context parallelism sizing, selective recompute, dispatcher choices, and patterns from DSV3 and Qwen3 experiments.

Practical guidance for training MoE vision-language models in Megatron Bridge, comparing FSDP and 3D-parallel approaches with lessons from Qwen3-VL and other multimodal experiments.

Skill for building a promotion case using strategies from 17 product leaders, covering visibility, advocacy, blocker diagnosis, and framing advancement as solving company problems

A UI design skill from the Generative Media Skills collection that generates high-fidelity mobile and web mockups using Atomic Design principles through muapi-cli, enabling rapid prototyping of professional interface designs.

Guides running Megatron-LM and Megatron Bridge training with mock or real data, covering correlation testing between MLM and Bridge loss curves, available recipes, and multi-GPU launch configuration.

Provides UI components for user-managed MCP (Model Context Protocol) server connections in assistant-ui chat interfaces, including OAuth authentication, server management, and connection workflows.

An NVIDIA VSS Blueprint skill for generating video analysis reports as part of the Video Search and Summarization platform.

Part of the Grafana skills collection, this skill provides administrative workflows for managing Grafana instances, covering dashboards, panels, PromQL, and the broader LGTM observability stack including Loki, Tempo, Prometheus, and Pyroscope.

Removes backgrounds from images to produce transparent PNGs, supporting portraits, product photos, pets, and group shots. Uses the Bria RMBG 2.0 model via fal.ai with no prompt required, processing in about 3 seconds at $0.01 per image.

Generates Obsidian Canvas, Excalidraw, and Mermaid diagrams from text descriptions using Claude Code for visual knowledge management.

A product management skill that guides AI agents through creating proto-personas to represent target user segments before conducting formal research.

AI agent skills focused on resume optimization, job applications, and career development for job seekers and professionals using Claude Code.

Deploys and operates the RTVI-CV-3D multi-view 3D tracking stack with per-camera DeepStream perception and BEV Fusion, supporting sample data, custom videos, and RTSP flows with automatic calibration chaining.

Part of the Grafana skills collection, this skill provides workflows for machine learning and AI capabilities within the Grafana observability platform, integrating with the LGTM stack (Loki, Grafana, Tempo, Mimir) for intelligent monitoring and analysis.

Enables MoE expert-parallel communication overlap in Megatron Bridge to hide dispatch/combine latency, supporting alltoall, DeepEP, and HybridEP flex dispatcher backends with optional delayed weight-gradient computation.

React Native and Expo development skill covering components, styling, animations, navigation, state management, networking, performance optimization, and deployment.

Operational guide for enabling hierarchical context parallelism in Megatron Bridge, using nested context-parallel process groups with a2a+p2p communication to scale CP beyond KV heads.

Indexes all shipped Megatron Bridge recipes and recommends the best starting training configuration based on model name/size, GPU count and type, and training goal (pretrain, SFT, or PEFT).

Part of the Grafana Skills collection providing guidance on private connectivity configuration for Grafana Cloud. Covers fleet management, cloud integrations, and secure agent connectivity within the LGTM observability stack.

Official NVIDIA-verified skill for tile-based GPU programming that guides AI agents through adding new CuTile kernels, performing cross-framework conversion, and optimizing performance. Part of the NVIDIA Agent Skills catalog with 110+ verified skills across 24 products, each signed and benchmarked.