2,883 件 ・ 関連度順
概要と使いどころ
Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.
日本語の概要は準備中です。原文の説明を表示しています。
Data framework for building LLM applications with RAG. Specializes in document ingestion (300+ connectors), indexing, and querying. Features vector indices, query engines, agents, and multi-modal support. Use for document Q&A, chatbots, knowledge retrieval, or building RAG pipelines. Best for data-centric LLM applications.
日本語の概要は準備中です。原文の説明を表示しています。
Convert a document to markdown and classify it with an LLM in a single pipeline.
日本語の概要は準備中です。原文の説明を表示しています。
Claude Code (code.claude.com) のデプロイ・CI/CD リファレンス。 Amazon Bedrock / Google Cloud Agent Platform (Vertex AI) / Microsoft Foundry / Claude Platform on AWS 経由、Claude apps gateway、LLM gateway (protocol / rollout)、 GitHub Actions (claude-code-action) / GitLab CI/CD / GitHub Enterprise Server (GHES)、 code review 自動化・ultrareview、Slack (Claude Tag)、headless mode、cloud environments、 devcontainer、self-hosted environments（runner 構成・identity 検証・本番デプロイ）。
Claude API (platform.claude.com) のプロンプトエンジニアリング・評価リファレンス。 prompt engineering, XML タグ, few-shot, system prompts, モデル別プロンプティング（Fable 5 / Opus 5 / Sonnet 5）。 evals, grading（LLM-graded 等）, jailbreak・prompt leak 対策, hallucination 低減, classification, ticket routing 等ユースケース実装。
窓際族物語の動画をクラウドを使わずフルローカルで生成するワークフロー（MiniMax H3 + ComfyUI）。ユーザーから「ローカルで動画作成して」「ローカルLLMで動画を作って」と指示されたとき、MiniMax H3での動画生成・修正を頼まれたときに必ず使用する。台本→Irodori-TTS/VOICEVOX音声→draw-things-cliキーフレーム一括生成→Qwen3-VL（Ollama）による画像検証→チャプター毎のH3動画生成→ffmpeg結合まで全工程をローカルで完結させる。
macOS の遅い回線や接続不良を netdoc コマンドで切り分け、根拠のある修復と再測定・復元まで行う。Web・動画・ダウンロード全体の遅さ、Wi-Fi・DNS・VPN・LLM 疎通の調査や改善に使う。
Token-slim mode — cut this session's token usage by routing every file read, code search, directory listing, and JSON dump through the token-slim MCP server (or its binary directly). Works in any repository or folder; the server itself is LLM-agnostic (any MCP client). Use when the user types /slim, or asks to save/reduce tokens, "トークン節約", "トークン削減", "コンテキスト節約", "token slim".
Harden LLM and agent integrations you operate: prompt injection, tool allowlists, secret leakage, and supply-chain of models. Defensive only.
日本語の概要は準備中です。原文の説明を表示しています。
Use when adding monitoring, metrics, logging, or tracing to Spring AI or LLM integration code. Covers token tracking, latency measurement, cost estimation, and prompt/response logging. Use when user mentions AI monitoring, token costs, or LLM observability.
日本語の概要は準備中です。原文の説明を表示しています。
Use when integrating LLMs, chat clients, embeddings, RAG pipelines, or AI agents into Spring Boot. Covers Spring AI ChatClient, prompt templates, embeddings, vector stores, and structured output. Use when user mentions Spring AI, LLM, ChatGPT, Claude, RAG, embeddings.
日本語の概要は準備中です。原文の説明を表示しています。
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.
日本語の概要は準備中です。原文の説明を表示しています。
Remove refusal behaviors from open-weight LLMs using OBLITERATUS — mechanistic interpretability techniques (diff-in-means, SVD, whitened SVD, LEACE, SAE decomposition, etc.) to excise guardrails while preserving reasoning. 9 CLI methods, 28 analysis modules, 116 model presets across 5 compute tiers, tournament evaluation, and telemetry-driven recommendations. Use when a user wants to uncensor, abliterate, or remove refusal from an LLM.
日本語の概要は準備中です。原文の説明を表示しています。
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.
日本語の概要は準備中です。原文の説明を表示しています。
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
日本語の概要は準備中です。原文の説明を表示しています。
Infraestrutura para IA — GPU, servidores MCP, gateway de LLM, pipeline de inferência, custo de token, cache semântico e observabilidade de modelo. Use quando o usuário disser "GPU", "MCP server", "LLM", "inferência", "modelo", "gateway de IA", "custo de token", "rate limit da API de IA", "self-host de modelo".
日本語の概要は準備中です。原文の説明を表示しています。
Build generative UI apps with OpenUI and OpenUI Lang — the token-efficient open standard for LLM-generated interfaces. Use when mentioning OpenUI, @openuidev, generative UI, streaming UI from LLMs, component libraries for AI, or replacing json-render/A2UI. Covers scaffolding, defineComponent, system prompts, the Renderer, and debugging OpenUI Lang output.
日本語の概要は準備中です。原文の説明を表示しています。
Anti-AI-slop design skill for greenfield pages, audits, redesigns, and design extraction from URLs or screenshots. Use when the user asks to build a new app or landing page, wants to redesign something, invokes Hallmark by name, or uses audit/redesign/study.
日本語の概要は準備中です。原文の説明を表示しています。
Design automated evaluation pipelines for LLM and agent systems — combining deterministic checks, statistical metrics, and LLM-as-judge scoring into repeatable, CI-integrated eval suites. Load when the user asks to set up automated evals, design an eval pipeline, integrate evals into CI/CD, create an eval suite, do eval-driven development, or says "automate my evals", "CI eval integration", "evaluation pipeline", "continuous evaluation", "monitoring eval quality", "set up regression testing for my agent". Sub-skill of eval-output orchestrator.
日本語の概要は準備中です。原文の説明を表示しています。
Build face recognition systems with InsightFace, ArcFace, enrollment pipelines, HDBSCAN clustering, and privacy-compliant architecture. Activate on: face recognition, face enrollment, face clustering, identity verification, facial search. NOT for: general object detection (computer-vision-pipeline), emotion analysis (ai-engineer).
日本語の概要は準備中です。原文の説明を表示しています。
Track real-time API cost accrual during LLM execution. Activate on 'cost tracking', 'token usage', 'API costs', 'budget monitoring', 'usage metrics'. NOT for cost estimation, pricing tiers, or billing systems.
日本語の概要は準備中です。原文の説明を表示しています。
Audit LLM token cost estimates against actual API usage. Activate on 'cost verification', 'token estimate accuracy', 'API cost audit', 'estimation variance'. NOT for pricing lookups, budget planning, or cost optimization strategies.
日本語の概要は準備中です。原文の説明を表示しています。
Produces a narrow-interval quality signal for a WinDAGs skill that a third party can verify without inspecting model weights or method internals. The mechanism is attribution-kNN over a tamper-evident outcome log (task→skills→accept/reject store), combined with a conformal-prediction calibration certificate that provides a formal coverage guarantee. Thompson/Beta sampling is explicitly rejected as a category error: it produces a selection signal for stationary i.i.d. rewards, not an attestation signal for context-dependent LLM output quality. The resulting bundle — outcome-log Merkle root + conformal threshold + optional TEE guardrail signature — constitutes a "narrow-interval trust" credential: verifiable, bounded, and internals-opaque.
日本語の概要は準備中です。原文の説明を表示しています。
Expert in background job processing with Bull/BullMQ (Redis), Celery, and cloud queues. Implements retries, scheduling, priority queues, and worker management. Use for async task processing, email campaigns, report generation, batch operations. Activate on "background job", "async task", "queue", "worker", "BullMQ", "Celery". NOT for real-time WebSocket communication, synchronous API calls, or simple setTimeout operations.
日本語の概要は準備中です。原文の説明を表示しています。