CLI 工具
命令行工具
training-llms-megatron
Orchestra-Research
Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.
optimizing-attention-flash
Orchestra-Research
Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.
lambda-labs-gpu-cloud
Orchestra-Research
Reserved and on-demand GPU cloud instances for ML training and inference. Use when you need dedicated GPU instances with simple SSH access, persistent filesystems, or high-performance multi-node clusters for large-scale training.
llama-cpp
Orchestra-Research
Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
evaluating-code-models
Orchestra-Research
Evaluates code generation models across HumanEval, MBPP, MultiPL-E, and 15+ benchmarks with pass@k metrics. Use when benchmarking code models, comparing coding abilities, testing multi-language support, or measuring code generation quality. Industry standard from BigCode Project used by HuggingFace leaderboards.
evaluating-llms-harness
Orchestra-Research
Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag). Use when benchmarking model quality, comparing models, reporting academic results, or tracking training progress. Industry standard used by EleutherAI, HuggingFace, and major labs. Supports HuggingFace, vLLM, APIs.
gguf-quantization
Orchestra-Research
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
distributed-llm-pretraining-torchtitan
Orchestra-Research
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
skill-scanner
getsentry
Scan agent skills for security issues. Use when asked to "scan a skill",
sred-work-summary
getsentry
Go back through the previous year of work and create a Notion doc that groups relevant links into projects that can then be documented as SRED projects.
sred-work-summary
getsentry
Go back through the previous year of work and create a Notion doc that groups relevant links into projects that can then be documented as SRED projects.
gh-review-requests
getsentry
Fetch unread GitHub notifications for open PRs where review is requested from a specified team or opened by a team member. Use when asked to "find PRs I need to review", "show my review requests", "what needs my review", "fetch GitHub review requests", or "check team review queue".
gh-review-requests
getsentry
Fetch unread GitHub notifications for open PRs where review is requested from a specified team or opened by a team member. Use when asked to "find PRs I need to review", "show my review requests", "what needs my review", "fetch GitHub review requests", or "check team review queue".
webapp-testing
snyk
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
snyk
Comprehensive PDF manipulation toolkit for extracting text and tables, creating new PDFs, merging/splitting documents, and handling forms. When Claude needs to fill in a PDF form or programmatically process, generate, or analyze PDF documents at scale.
agency-docs-updater
glebis
End-to-end pipeline for publishing Claude Code lab meetings. Automatically finds/creates Fathom transcript, downloads video, uploads to YouTube, generates fact-checked Russian summary, creates MDX documentation, and pushes to agency-docs for Vercel deployment. Single invocation replaces 5+ manual steps.
gmgn-track
GMGNAI
Get real-time crypto buy/sell activity from Smart Money wallets, KOL influencer wallets, and personally followed wallets via GMGN API — alpha signals, whale tracking, meme token copy-trading ideas on Solana, BSC, Base, or Ethereum. Also query which tokens a wallet has followed (bookmarked) on GMGN. Use when user asks what smart money or KOLs are buying, wants whale alerts, on-chain alpha, copy-trade signals, or wants to check a wallet's followed tokens. (For a specific wallet address's portfolio, use gmgn-portfolio.)
gmgn-portfolio
GMGNAI
Analyze any crypto wallet by address — holdings, realized/unrealized P&L, win rate, trading history, performance stats, specific token balance, and tokens created by a developer wallet (with ATH market cap and DEX graduation status) via GMGN API on Solana, BSC, Base, or Ethereum. Use when user asks about a wallet's holdings, P&L, win rate, what tokens a dev has launched, the highest ATH token a dev ever created, or wants a wallet report to decide whether to copy-trade or follow.
gmgn-swap
GMGNAI
"[FINANCIAL EXECUTION] Buy and sell meme coins and crypto tokens on Solana, BSC, Base, or Ethereum — single swap, multi-wallet batch trading, limit orders, stop loss, take profit, trailing stop loss, trailing take profit via GMGN API. Requires explicit user confirmation. Use when user asks to buy, sell, or swap a token, trade from multiple wallets, set a limit order, stop loss, take profit, or check order status."
claude-ecosystem
majiayu000
"Claude Code ecosystem expertise. Modules: CLI tool (setup, slash commands, MCP servers, hooks, plugins, CI/CD), extensibility (agents, skills, output styles creation), CLAUDE.md (project instructions, optimization). Actions: configure, troubleshoot, create, deploy, integrate, optimize Claude Code. Keywords: Claude Code, Anthropic, CLI tool, slash command, MCP server, Agent Skill, hook, plugin, CI/CD, enterprise, CLAUDE.md, agentic coding, agent, skill, output-style, SKILL.md, subagent, Task tool, project instructions, token optimization. Use when: learning Claude Code features, configuring settings, creating skills/agents/hooks, setting up MCP servers, troubleshooting issues, CI/CD integration, initializing or optimizing CLAUDE.md files."
state-directory-manager
majiayu000
Manage persistent state directories for bash scripts
state-directory-manager
majiayu000
Manage persistent state directories for bash scripts
codex-delegator
majiayu000
Automatically delegate complex, logic-intensive tasks to OpenAI Codex CLI via codex exec --full-auto. Claude Code uses this skill to invoke Codex for complex backend logic, intricate algorithms, or persistent bugs. Enables seamless AI-to-AI collaboration where Claude Code analyzes and Codex executes.
buildkite
majiayu000
Use this skill when working with Buildkite CI/CD pipelines, agents, builds, test analytics, or package registries. Covers pipeline configuration, YAML syntax, agent setup, build steps, plugins, and API usage.