1,235 件 ・ 関連度順
概要と使いどころ
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.
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。
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.
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 guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
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 guidance for Fully Sharded Data Parallel training with PyTorch FSDP - parameter sharding, mixed precision, CPU offloading, FSDP2
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。
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.
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。
Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.
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。
React composition patterns that scale. Use when refactoring components with boolean prop proliferation, building flexible component libraries, or designing reusable APIs. Triggers on tasks involving compound components, render props, context providers, or component architecture. Includes React 19 API changes.
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 cognee is slow, expensive, or hitting rate limits — speeding up or throttling ingestion (remember/cognify batching, chunk size, LLM and embedding rate limits, per-stage models), cutting read latency, estimating cost before ingesting, running work in the background, or planning how to scale a deployment.
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。
Publish the user's own gbrain over MCP so other devices, desktop apps and cloud agents can reach it. `gbrain mcp expose` installs and signs in Tailscale, publishes the running `gbrain serve --http` on the tailnet (HTTPS, tailnet-only by default; Funnel only when a client lives in a vendor cloud), keeps the server alive as a user service, and hands back the MCP URL. Then register one least-privilege client per consumer, choose native OAuth/PKCE or a private machine handoff for that client, and verify the connection inside the intended harness.
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。
End-to-end discipline for turning any large data source (audio libraries, email takeouts, document corpora, chat exports, API dumps) into brain pages at scale. The lifecycle spine: SCHEMA → ACCESS → TRIAL → EVALUATE → IMPROVE → CODIFY → TEST → SKILLIFY → BULK → MONITOR. State is tracked in a durable JSON manifest (see MANIFEST-PATTERN.md) so any crash, session boundary, or subagent fan-out resumes from ground truth instead of memory.
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。
Senior Project Manager for enterprise software, SaaS, and digital transformation projects. Specializes in portfolio management, quantitative risk analysis, resource optimization, stakeholder alignment, and executive reporting. Uses advanced methodologies including EMV analysis, Monte Carlo simulation, WSJF prioritization, and multi-dimensional health scoring. Use when a user needs help with project plans, project status reports, risk assessments, resource allocation, project roadmaps, milestone tracking, team capacity planning, portfolio health reviews, program management, or executive-level project reporting — especially for enterprise-scale initiatives with multiple workstreams, complex dependencies, or multi-million dollar budgets.
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。
When the user wants to audit, review, or diagnose SEO issues on their site. Also use when the user mentions "SEO audit," "technical SEO," "why am I not ranking," "SEO issues," "on-page SEO," "meta tags review," or "SEO health check." For building pages at scale to target keywords, see programmatic-seo. For adding structured data, see schema-markup.
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。
When the user needs to generate, iterate, or scale ad creative for paid advertising. Use when they say 'write ad copy,' 'generate headlines,' 'create ad variations,' 'bulk creative,' 'iterate on ads,' 'ad copy validation,' 'RSA headlines,' 'Meta ad copy,' 'LinkedIn ad,' or 'creative testing.' This is pure creative production — distinct from paid-ads (campaign strategy). Use ad-creative when you need the copy, not the campaign plan.
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。
Cross-functional organizational health check combining signals from all C-suite roles. Scores 8 dimensions on a traffic-light scale with drill-down recommendations. Use when assessing overall company health, preparing for board reviews, identifying at-risk functions, or when user mentions org health, health check, or health dashboard.
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 a Head of People Ops, BizOps lead, or Internal Communications owner needs to draft and sequence an internal-only change-management communication — a re-org announcement, a tool rollout, a policy change, a leadership transition, a layoff, an acquisition close, or an internal product launch — and the audience is employees (not customers). Pairs Prosci ADKAR and Kotter's 8-step change model with deterministic stdlib-only Python tools to produce a sequenced touchpoint calendar, a Kotter-compliant primary announcement, an audience-segmented FAQ, and manager cascade talking points; industry-tuned via --profile {tech-startup, scaleup, enterprise, public-company, non-profit}. Triggers on "all-hands announcement", "change comms", "rollout comms", "re-org announcement", "manager talking points", "layoff comms".
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。
Analyze user feedback data to identify segments with sentiment scores, JTBD, and product satisfaction insights. Use when analyzing user feedback at scale, running sentiment analysis on reviews or surveys, or identifying satisfaction patterns.
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。
Review a PR against the Pascal architectural rules — package boundaries (core/viewer/editor/nodes), the registry-driven composition model (def.geometry / def.renderer / def.system), legacy-dispatch regressions, the slots + world-scale-UV convention for new nodes/geometry, hook hygiene (useEditor/useScene/useViewer), and selector performance. Use when the user asks to review a PR, audit a branch, or check that changes respect the codebase's architecture.
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。
Assess whether furniture fits in a measured Pascal room or layout. Use this skill for sofa, table, bed, cabinet, appliance, staging, placement, collision, clearance, or rotated-footprint questions. Produce a tool-backed spatial report that distinguishes footprint fit from unsupported height, door-swing, assembly, and delivery-route claims, and return insufficient evidence when dimensions or scale are missing.
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, deploy and operate Kortix Apps through the pre-authenticated CLI or SDK: static sites and SPAs, Dockerfile or OCI servers, workers, and stateful backends (a database, a Convex backend) with persistent volumes, snapshots and built-in Kortix sign-in. Use when the user asks to publish, host, deploy, preview, inspect, scale, wake, suspend, roll back, debug, or remove an App; asks for a database, an API, a backend, a data model, auth, or 'store this'; asks for an internal tool, business app, CRM, tracker, dashboard, portal, or any app the team logs in to; before writing any Convex code; or asks what an App costs or how it runs.
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。
Field-tested methodology and concrete recipes for training and operating large-scale LLM/VLM/multi-modal models end to end - choosing and benchmarking accelerators, storage and network; SLURM/Kubernetes orchestration; maximizing training throughput and fitting models in memory; diagnosing and surviving training instabilities, NaN/Inf, and hardware/job failures; checkpointing and fault tolerance; inference performance and memory; debugging multi-node/ multi-GPU hangs; and writing/running tests. Use when the user is training or fine-tuning large models, hits low TFLOPS/MFU, OOM, slow dataloading, a loss spike/divergence, a NCCL/InfiniBand or multi-node hang, node/GPU failures, checkpoint or preemption problems, storage/network bottlenecks, or needs to pick GPUs/cloud/file-systems or size inference latency/throughput. Distilled from "Machine Learning Engineering", the latest version of which can be found at https://github.com/stas00/ml-engineering The latest SKILL.md version can be found at https://github.com/stas00/ml-engineering/blob/master/skills/ml-engineering/SKILL.md
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。
Scaffolding that helps a writer build a story they already want to tell — fills in structure, hidden wound, theme, and prose from the writer's own material across seven narrative layers, deriving the spine from their notes as proposals they ratify. Bans generic AI patterns; scales short story to multi-book series. USE WHEN write story, fiction, novel, chapter, story bible, character arc, plot outline, creative writing, worldbuilding, draft story, help me write my story, develop my novel, layer my story, build out my book, I have notes for a story. NOT FOR narrative summaries of real content.
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。
Defensive threat modeling and risk management for your own estate — map where sensitive data lives across your asset graph, run compromise scenarios (what a hacked asset exposes, its blast radius, how you'd respond), and maintain a persistent risk register with likelihood×impact scoring, owners, mitigations, and review cadence via a deterministic CLI. All real data stays in your private USER tree; this skill is code only. USE WHEN threat model, threat modeling, risk register, risk assessment, what if X got hacked, compromise scenario, blast radius, sensitive data map, where is our sensitive data, data classification, risk review, add a risk, accept a risk, security risk posture. NOT FOR active pentesting or exploitation (use an offensive-security skill), world-scale futures stress-testing of ideas (use WorldThreatModel), or executing incident response (use your incident-response runbooks — this skill plans them).
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。
Watch a video (URL or local path). Downloads with yt-dlp, extracts auto-scaled frames with ffmpeg, pulls the transcript from captions (or local WhisperX / cloud Whisper fallback), and hands the result to the agent so it can answer questions about what's in the video. With a Gemini API key, Google's agentic video model watches the full video instead.
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 creating, extending, debugging or publishing a Lightweight Charts™ plugin — a custom series, a series primitive or a pane primitive — or when turning a chart idea ("I want the chart to show/draw X") into a working plugin. Covers choosing the plugin type, scaffolding with create-lwc-plugin, building on @tradingview/lwc-toolkit instead of hand-rolled helpers, which of the official plugin packages to read as the reference implementation, where the docs are, and the autoscale, whitespace, visible-range, hit-test and conflation mistakes every plugin author makes once. Reach for it whenever the user mentions a custom series, primitive, renderer, drawing on the chart canvas, lwc-plugin, plugin-examples, the toolkit, or the plugin catalog — even if they never say the word "plugin".
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。