Claude Skills by anubhavg-icpl

Expert in the Tetris design system - Classic block-game inspired design with playful colors, bold display fonts, and compact, high-energy layouts. Use when building UI components, applying design tokens, or implementing visual styles for tetris.

Run oobabooga's textgen with multiple backends, OpenAI/Anthropic-compatible API, characters, training tab, and portable installer. Use when deploying, running, or configuring local LLM inference with text generation webui.

Deploy HuggingFace TGI in Docker with sharding, AWQ/GPTQ/EETQ/bitsandbytes quantization, and the OpenAI-compatible Messages API. Use when deploying, running, or configuring local LLM inference with tgi huggingface.

Expert in the The Verge design system - Tech editorial media. Acid-mint and ultraviolet accents, Manuka display, rave-flyer story tiles. Use when building UI components, applying design tokens, or implementing visual styles for theverge.

Expert in TimescaleDB hypertables, continuous aggregates, and compression. Use when working with timescaledb for data processing, streaming, or analytics.

Expert in Tinybird real-time analytics APIs on managed ClickHouse. Use when working with tinybird for data processing, streaming, or analytics.

TJ Holowaychuk — Express, Koa, Mocha, Commander; minimal cores, prolific output, small modules that compose. Use when you want code review, architecture advice, or opinions in the style of tj holowaychuk.

Expert in the Together AI design system - Open-source AI infrastructure. Technical, blueprint-style design. Use when building UI components, applying design tokens, or implementing visual styles for together ai.

RFC compliance validator for Token Administration Console implementations - OAuth 2.0, Bearer Token, Token Introspection. Use when you need help with token administration console validator.

Compute token counts, inference latency, and quality ceiling for Emu3-style next-token generation and pick between Emu3-family and diffusion. Use when you need help with token gen cost analyzer.

Train tokenizers from scratch with HF tokenizers — BPE / SentencePiece / WordPiece — extend vocab for new languages or code, and add chat / special tokens. Use when creating, converting, or publishing model files with tokenizer engineering.

Pick a tokenization approach for a new language model or text pipeline. Use when you need help with tokenizer picker.

Pick between Chameleon-style early fusion (shared-vocab tokenizer) and LLaVA-style late fusion (adapter on frozen LLM) for a VLM project. Use when you need help with tokenizer vs adapter picker.

Audit a multi-agent system that claims \"emergent coordination.\" Separates real ToM-enabled coordination from prompt-dressed illusion with control conditions, statistical tests, and complementarity measurement. Use when you need help with tom auditor.

Tony Stark / Iron Man coding assistant - genius billionaire playboy philanthropist with unmatched tech skills and an ego to match. Use when you need help with tony stark.

Audit a tool definition (name + description + JSON Schema + executor outline) for loop fitness before it ships to an LLM. Use when you need help with tool interface reviewer.

Build a production tool catalog and registry with JSON Schema validation, parallel dispatch, and observability. Use when you need help with tool registry.

Audit a tool registry against production design rules for names, descriptions, parameters, and shape. Can run in CI on every tool-registry change. Use when you need help with tool schema linter.

Expert in tool use, function calling, and MCP — the interfaces between AI and the real world, from the AI Engineering from Scratch curriculum. Use when you need help with tools and protocols.

Pick LDA or BERTopic for a corpus. Specify library, knobs, evaluation. Use when you need help with topic picker.

Pick a multi-agent debate topology (star / chain / tree / graph), an N of agents, a heterogeneity profile, and a round bound for a given task. Use when you need help with topology picker.

Linus Torvalds — taste, no special cases, brutal directness, \"talk is cheap, show me the code. Use when you want code review, architecture advice, or opinions in the style of torvalds.

tracing. Use when you need help with tracing.

|. Use when you need help with trading analysis dashboard template.

Estimate (N, D, hours, GPU count) for a new transformer training run given compute budget and deployment constraints. Use when you need help with training budget estimator.

Review a transformer block implementation against 2026 defaults and flag drift. Use when you need help with transformer block reviewer.

Review a transformer-from-scratch implementation against the 13 Phase 7 lessons. Use when you need help with transformer review.

Expert in transformer architectures from self-attention to modern variants, from the AI Engineering from Scratch curriculum. Use when you need help with transformers deep dive.

Run durable AI background jobs and agents with Trigger.dev v3 — no timeouts, full observability. Use when building AI applications with trigger dev.

Review a proposed agent detector stack (kill switch, circuit breakers, canary tokens) and flag missing tripwires before the first autonomous run. Use when building UI with tripwire design language and visual style.

HuggingFace TRL — SFTTrainer, DPOTrainer, PPOTrainer, GRPOTrainer, RewardTrainer. Use when fine-tuning, training, or adapting language models with trl techniques.

Expert in tRPC, Drizzle ORM, and Zod for end-to-end type-safe full-stack TypeScript applications. Use when you need help with trpc fullstack.

Decide whether Blackwell + TensorRT-LLM + Dynamo is worth the NVIDIA-lock for a given workload and budget. Use when you need help with trtllm blackwell advisor.

Pick TTS model, voice, text-normalization scope, and evaluation plan for a given language, style, and latency target. Use when you need help with tts designer.

Deep expertise in Turbopuffer — object-storage-first vector + full-text search, three-tier caching, and ~$0.02/GB cold storage economics. Use when implementing vector search, embeddings storage, or similarity queries with turbopuffer.

Expert in Turborepo 2 monorepo orchestration, task pipelines, and remote caching. Use when building web applications with turborepo.

Expert in Turborepo, monorepo architecture, workspace management, and build optimization. Use when designing system architecture or making technical decisions about turborepo monorepo.

|. Use when you need help with tweaks.

Design a Transfusion / MMDiT-style two-loss training setup (NTP on one modality, diffusion on another) with loss weights, mask design, and schedule. Use when you need help with two loss trainer designer.

Expert in TypeORM with TypeScript, entities, repositories, migrations, query building, and production best practices. Use when you need deep expertise in typeorm.

Advanced TypeScript patterns including generics, conditional types, mapped types, template literals, and type guards. Use when you need help with typescript advanced.

Production-ready TypeScript coding standards enforcing type safety, modern patterns, and maintainability. Use when enforcing typescript coding conventions and style rules.

Production-ready TypeScript Node.js project structure architect - validates and scaffolds enterprise-grade Node.js/Bun applications with monorepo patterns. Use when scaffolding, structuring, or architecting typescript node projects.

Expert in the Uber design system - Mobility platform. Bold black and white, tight type, urban energy. Use when building UI components, applying design tokens, or implementing visual styles for uber.

ui-designer. Use when you need help with ui designer.

Pick between Show-o / Transfusion / Emu3 / Janus-Pro families for a product that needs both multimodal understanding and generation with open weights. Use when you need help with unified gen model picker.

unity-developer. Use when developing games with unity.

Universal performance optimization best practices for all languages, frameworks, and stacks. Comprehensive guidance for frontend, backend, database, and infrastructure optimization with actionable checklists and real-world examples. Use when you need help with universal performance optimization.

Universal rules-based AI assistant that loads and follows project-specific coding standards, workflows, and best practices across all AI platforms (Amazon Q, Claude Code, Copilot, Gemini, Aider, Codex). Use when you need help with universal rules engine.

unreal-engine. Use when developing games with unreal engine.