全部能力

Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory),…

Philosophy-aware PR reviews checking alignment with amplihack principles. Use when reviewing PRs to ensure ruthless simplicity, modular design, and zero-BS implementation. Suggests simplifications, identifies over-engineering, verifies brick module structure. Posts detailed, constructive review comments with specific file:line references.

React web development with hooks, React Query, Zustand

Creates, refines, and validates Agent Skills following the open standard at agentskills.io, Claude Code extensions, and Anthropic best practices. Use when…

Upgrade any Pulumi provider to a newer version and reconcile the resulting diff. Use when users want to upgrade or update a provider (including editing package.json, requirements.txt, pyproject.toml, go.mod, or Pulumi.yaml to bump a provider SDK), check for breaking changes before or during an upgrade, fix resources that broke after a provider upgrade, or resolve unexpected replacements, creates, or deletes in a post-upgrade preview. Applies to all providers (aws, azure-native, gcp, kubernetes, aws-native, cloudflare, datadog, etc.) — not just Tier 1. Do NOT use for querying which stacks use what package versions; use skill `package-usage` for cross-stack audits. Do NOT use for general infrastructure tasks.

Systematic step-by-step reasoning with revision and branching. Use for complex problems, multi-stage analysis, design planning, problem decomposition, or…

Use when asked to "shape up", "run a shaping session", "set an appetite", "scope a project without estimates", "betting table", or "ship in fixed cycles".…

Technical SEO - robots.txt, sitemap, meta tags, Core Web Vitals

SEO keyword clustering with topic organization, content hub architecture, internal linking strategies. Use for content strategy, keyword research, pillar page…

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model…

Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation…

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…

How to choose and configure data sources for MapLibre GL JS — rendering your own data without tiles, hosted tile services, serverless PMTiles, self-hosted tile…

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

This skill should be used when the user asks to 'place a CEX order', 'trade on centralized exchange', 'buy BTC on CEX', 'sell ETH futures', 'open a long…

| Nuxt 4 server-side development with Nitro: API routes, server middleware, database integration, and backend patterns. Use when: creating server API routes,…

LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring…

Python video composition with moviepy 2.x — overlaying deterministic text on AI-generated video (LTX-2, SadTalker), compositing clips, single-file build.py…

Expert guidance for fine-tuning LLMs with LLaMA-Factory - WebUI no-code, 100+ models, 2/3/4/5/6/8-bit QLoRA, multimodal support

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