Agent Skills Directory

Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when fine-tuning large models (7B-70B) with limited GPU memory, when you need to train <1% of parameters with minimal accuracy loss, or for multi-adapter serving. HuggingFace's official library integrated with transformers ecosystem.

Open-source AI observability platform for LLM tracing, evaluation, and monitoring. Use when debugging LLM applications with detailed traces, running evaluations on datasets, or monitoring production AI systems with real-time insights.

Managed vector database for production AI applications. Fully managed, auto-scaling, with hybrid search (dense + sparse), metadata filtering, and namespaces. Low latency (<100ms p95). Use for production RAG, recommendation systems, or semantic search at scale. Best for serverless, managed infrastructure.

Generates conference presentation slides (Beamer LaTeX PDF and editable PPTX) from a compiled paper with speaker notes and talk script. Use when preparing oral talks, spotlight presentations, or invited talks for ML and systems conferences.

Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+ TPR, <1% FPR. Fast (<2ms GPU). Multilingual (8 languages). Deploy with HuggingFace or batch processing for RAG security.

Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you need DTensor-based sharding with DeviceMesh.

High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.

Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention framework. Use when conducting causal tracing, activation patching, interchange intervention training, or testing causal hypotheses about model behavior.

High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.

Quantizes LLMs to 8-bit or 4-bit for 50-75% memory reduction with minimal accuracy loss. Use when GPU memory is limited, need to fit larger models, or want faster inference. Supports INT8, NF4, FP4 formats, QLoRA training, and 8-bit optimizers. Works with HuggingFace Transformers.

Scalable data processing for ML workloads. Streaming execution across CPU/GPU, supports Parquet/CSV/JSON/images. Integrates with Ray Train, PyTorch, TensorFlow. Scales from single machine to 100s of nodes. Use for batch inference, data preprocessing, multi-modal data loading, or distributed ETL pipelines.

Distributed training orchestration across clusters. Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nodes. Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling. Use when training massive models across multiple machines or running distributed hyperparameter sweeps.

RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.

Foundation model for image segmentation with zero-shot transfer. Use when you need to segment any object in images using points, boxes, or masks as prompts, or automatically generate all object masks in an image.

Framework for state-of-the-art sentence, text, and image embeddings. Provides 5000+ pre-trained models for semantic similarity, clustering, and retrieval. Supports multilingual, domain-specific, and multimodal models. Use for generating embeddings for RAG, semantic search, or similarity tasks. Best for production embedding generation.

Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory), deterministic vocabulary. Used by T5, ALBERT, XLNet, mBART. Train on raw text without pre-tokenization. Use when you need multilingual support, CJK languages, or reproducible tokenization.

Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

Multi-cloud orchestration for ML workloads with automatic cost optimization. Use when you need to run training or batch jobs across multiple clouds, leverage spot instances with auto-recovery, or optimize GPU costs across providers.

Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling.

Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.

Accelerate LLM inference using speculative decoding, Medusa multiple heads, and lookahead decoding techniques. Use when optimizing inference speed (1.5-3.6× speedup), reducing latency for real-time applications, or deploying models with limited compute. Covers draft models, tree-based attention, Jacobi iteration, parallel token generation, and production deployment strategies.

State-of-the-art text-to-image generation with Stable Diffusion models via HuggingFace Diffusers. Use when generating images from text prompts, performing image-to-image translation, inpainting, or building custom diffusion pipelines.

Comprehensive guide for writing systems papers targeting OSDI, SOSP, ASPLOS, NSDI, and EuroSys. Provides paragraph-level structural blueprints, writing patterns, venue-specific checklists, reviewer guidelines, LaTeX templates, and conference deadlines. Use this skill for all systems conference paper writing.

Visualize training metrics, debug models with histograms, compare experiments, visualize model graphs, and profile performance with TensorBoard - Google's ML visualization toolkit

Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.

Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.

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.

Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.

Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization

Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.

Track ML experiments with automatic logging, visualize training in real-time, optimize hyperparameters with sweeps, and manage model registry with W&B - collaborative MLOps platform

OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.

Create cinematic product videos from shot recipe cards, a validated template, and code/audio assets (Remotion + real page screenshots + 2.5D camera moves + beat-synced cuts + sound design). Use when the user asks to turn a frontend project or webpage into a product video, says "use video-shotcraft to make a video/promo", names the Ink Press template or asks to reproduce its effect, or wants a single shot card's motion. 用镜头配方卡 + 已验收模板 + 代码/音频资产制作电影感产品视频（Remotion + 真实页面截图 + 2.5D 运镜 + 节奏卡点 + 声音设计）。当用户要求"用 video-shotcraft 做视频/宣传片"、把前端项目/网页做成产品视频、点名 Ink Press 模板或要求复刻模板片效果，或要用镜头卡做单个动效镜头时使用。

Typeset professional documents with Kami templates: resumes, one-pagers, white papers, letters, portfolios, and slide decks. Use when asked to 做 PDF / 排版 / 简历 / 一页纸 / PPT / slides, or to create a Kami landing page. Not for auditing or restyling an existing product site.

Generates Angular 17+ standalone components, configures advanced routing with lazy loading and guards, implements NgRx state management, applies RxJS patterns, and optimizes bundle performance. Use when building Angular 17+ applications with standalone components or signals, setting up NgRx stores, establishing RxJS reactive patterns, performance tuning, or writing Angular tests for enterprise apps.

Use when designing REST or GraphQL APIs, creating OpenAPI specifications, or planning API architecture. Invoke for resource modeling, versioning strategies, pagination patterns, error handling standards.

Use when designing new high-level system architecture, reviewing existing designs, or making architectural decisions. Invoke to create architecture diagrams, write Architecture Decision Records (ADRs), evaluate technology trade-offs, design component interactions, and plan for scalability. Use for system design, architecture review, microservices structuring, ADR authoring, scalability planning, and infrastructure pattern selection — distinct from code-level design patterns or database-only design tasks.

Integrates with Atlassian products to manage project tracking and documentation via MCP protocol. Use when querying Jira issues with JQL filters, creating and updating tickets with custom fields, searching or editing Confluence pages with CQL, managing sprints and backlogs, setting up MCP server authentication, syncing documentation, or debugging Atlassian API integrations.

Designs chaos experiments, creates failure injection frameworks, and facilitates game day exercises for distributed systems — producing runbooks, experiment manifests, rollback procedures, and post-mortem templates. Use when designing chaos experiments, implementing failure injection frameworks, or conducting game day exercises. Invoke for chaos experiments, resilience testing, blast radius control, game days, antifragile systems, fault injection, Chaos Monkey, Litmus Chaos.

Use when building CLI tools, implementing argument parsing, or adding interactive prompts. Invoke for parsing flags and subcommands, displaying progress bars and spinners, generating bash/zsh/fish completion scripts, CLI design, shell completions, and cross-platform terminal applications using commander, click, typer, or cobra.

Designs cloud architectures, creates migration plans, generates cost optimization recommendations, and produces disaster recovery strategies across AWS, Azure, and GCP. Use when designing cloud architectures, planning migrations, or optimizing multi-cloud deployments. Invoke for Well-Architected Framework, cost optimization, disaster recovery, landing zones, security architecture, serverless design.

Generates, formats, and validates technical documentation — including docstrings, OpenAPI/Swagger specs, JSDoc annotations, doc portals, and user guides. Use when adding docstrings to functions or classes, creating API documentation, building documentation sites, or writing tutorials and user guides. Invoke for OpenAPI/Swagger specs, JSDoc, doc portals, getting started guides.

Analyzes code diffs and files to identify bugs, security vulnerabilities (SQL injection, XSS, insecure deserialization), code smells, N+1 queries, naming issues, and architectural concerns, then produces a structured review report with prioritized, actionable feedback. Use when reviewing pull requests, conducting code quality audits, identifying refactoring opportunities, or checking for security issues. Invoke for PR reviews, code quality checks, refactoring suggestions, review code, code quality. Complements specialized skills (security-reviewer, test-master) by providing broad-scope review across correctness, performance, maintainability, and test coverage in a single pass.

Writes, optimizes, and debugs C++ applications using modern C++20/23 features, template metaprogramming, and high-performance systems techniques. Use when building or refactoring C++ code requiring concepts, ranges, coroutines, SIMD optimization, or careful memory management — or when addressing performance bottlenecks, concurrency issues, and build system configuration with CMake.

Use when building C# applications with .NET 8+, ASP.NET Core APIs, or Blazor web apps. Builds REST APIs using minimal or controller-based routing, configures database access with Entity Framework Core, implements async patterns and cancellation, structures applications with CQRS via MediatR, and scaffolds Blazor components with state management. Invoke for C#, .NET, ASP.NET Core, Blazor, Entity Framework, EF Core, Minimal API, MAUI, SignalR.

Optimizes database queries and improves performance across PostgreSQL and MySQL systems. Use when investigating slow queries, analyzing execution plans, or optimizing database performance. Invoke for index design, query rewrites, configuration tuning, partitioning strategies, lock contention resolution.