Loading...
Loading...
Found 35 Skills
Writes, refactors, and evaluates prompts for LLMs — generating optimized prompt templates, structured output schemas, evaluation rubrics, and test suites. Use when designing prompts for new LLM applications, refactoring existing prompts for better accuracy or token efficiency, implementing chain-of-thought or few-shot learning, creating system prompts with personas and guardrails, building JSON/function-calling schemas, or developing prompt evaluation frameworks to measure and improve model performance.
AI/LLM: Use when crafting system prompts, optimizing LLM outputs, or improving agent instructions. NOT for general coding.
Claude Code skill (trtllm-agent-toolkit): implement or extend TensorRT-LLM AutoDeploy fusion transforms under transform/library/ in a TensorRT-LLM checkout. Prefer existing kernels and custom ops; use Triton only when no viable existing-kernel path exists. Use ad-graph-dump for AD_DUMP_GRAPHS_DIR workflows. Covers TRT-LLM paths, registry, default.yaml registration, graph validation, tests, and a review checklist — without prescribing profiling tools or throughput targets.
Integrate TileGym kernels into Hugging Face `transformers` models by replacing the library's submodule(s) and certain class(es)' implementations, and patching certain class(es)' init/forward/load weight methods prior to instantiating models. Used when the user requires integrating TileGym kernels into `transformers` models.
Retrieval-Augmented Generation patterns including chunking, embeddings, vector stores, and retrieval optimizationUse when "rag, retrieval augmented, vector search, embeddings, semantic search, document qa, rag, retrieval, embeddings, vector, search, llm" mentioned.
Intelligent system governor that continuously shadow-tests APIs for performance while enforcing strict financial and security guardrails against runaway costs.
Add EAGLE-3 or draft-model speculative decoding to a Jetson vLLM server when TPOT is the bottleneck.
Expert skill for using TileKernels, a library of optimized GPU kernels for LLM operations (MoE routing, quantization, transpose, engram gating, Manifold HyperConnection) built with TileLang.
Generate a complete, ready-to-paste prompt engineered for a specific target model (Claude Fable 5.1 / Opus 5.5 / Sonnet 5 / Haiku 4.5, GPT-5.x) and harness (Claude Code CLAUDE.md or skill, Codex AGENTS.md, pi, Amp, eve, AI Gateway, Agent SDK, raw API, chat UI), using per-model and per-harness reference guides. Trigger on "metaprompt", "write me a prompt for", "generate a system prompt", "prompt engineer this", "optimize this prompt for GPT-5 / Opus", "hazme un prompt para", "escríbeme un system prompt", "mejora este prompt", "prompt para Codex / Claude Code". Not for executing the task the prompt describes, not for sending a prompt to another model (use askcodex), and not for building a runnable Claude Agent SDK agent (use agent-sdk-wizard).
Quickly test and compare LLM models via OpenRouter. Find the fastest/cheapest model, compare response quality. Trigger words: openrouter, test model, compare models, find fastest model, find cheapest model
Use this skill when you writing commands, hooks, skills for Agent, or prompts for sub agents or any other LLM interaction, including optimizing prompts, improving LLM outputs, or designing production prompt templates.