Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse...

Vision-language pre-training framework bridging frozen image encoders and LLMs. Use when you need image captioning, visual question answering, image-text retrieval, or multimodal chat with...

Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models...

Extract structured data from LLM responses with Pydantic validation, retry failed extractions automatically, parse complex JSON with type safety, and stream partial results with Instructor -...

Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training

Guarantee valid JSON/XML/code structure during generation, use Pydantic models for type-safe outputs, support local models (Transformers, vLLM), and maximize inference speed with Outlines -...

Use when creating new skills, editing existing skills, or verifying skills work before deployment

Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention

Expert guidance for Fully Sharded Data Parallel training with PyTorch FSDP - parameter sharding, mixed precision, CPU offloading, FSDP2

Raptor turns Claude Code into a general-purpose AI offensive/defensive security agent. By using Claude.md and creating rules, sub-agents, and skills, and orchestrating security tool usage, we...

# Liberty House Club **A Parallel Binance Chain to Enable Smart Contracts** _NOTE: This document is under development. Please check regularly for updates!_ ## Table of Contents -...

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate...

Application security testing coordinator for common vulnerability patterns including XSS, injection flaws, and client-side security issues. Orchestrates specialized testing agents to identify and...

Red-green-refactor development methodology requiring verified test coverage. Use for feature implementation, bugfixes, refactoring, or any behavior changes where tests must prove correctness.

Collaborative design exploration that refines ideas into validated specs through iterative questioning. Use before any creative work including creating features, building components, adding...

Git branch completion workflow. Use when implementation is complete, tests pass, and a feature branch needs to be integrated via merge, pull request, or cleanup.

Meta-skill enforcing skill discovery and invocation discipline through mandatory workflows. Use when starting any conversation to check for relevant skills before any response, ensuring...

IDA Pro 逆向分析。通过 IDAPython 脚本获取反汇编、反编译、字符串、导入表、交叉引用等信息。

Offensive AI security testing and exploitation framework. Systematically tests LLM applications for OWASP Top 10 vulnerabilities including prompt injection, model extraction, data poisoning, and...

Root cause analysis for debugging. Use when bugs, test failures, or unexpected behavior have non-obvious causes, or after multiple fix attempts have failed.