649 件 ・ 関連度順
概要と使いどころ
Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.
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 the user asks about finding the best, top, or recommended model for a task, wants to know what AI model to use, or wants to compare models by benchmark scores. Triggers on: "best model for X", "what model should I use for", "top models for [task]", "which model runs on my laptop/machine/device", "recommend a model for", "what LLM should I use for", "compare models for", "what's state of the art for", or any question about choosing an AI model for a specific use case. Always use this skill when the user wants model recommendations or comparisons, even if they don't explicitly mention HuggingFace or benchmarks.
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。
Evil Twin / KARMA / Mana access point methodology — rogue AP construction with hostapd-mana / wifiphisher / airgeddon, KARMA universal probe response, Mana selective probe response, captive portal phishing, deauth-driven client coercion to attacker AP, MAC randomization defeat via PNL leak analysis, post-association MITM (DNS, ARP, transparent proxy), credential capture for portal/web/SMB, and detection-evasion tactics. Use to coerce client devices onto an attacker-controlled AP, intercept their traffic, harvest credentials, or deliver payloads via captive portal.
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。
LoRaWAN and sub-GHz (433 / 868 / 915 MHz) attack methodology — LoRaWAN ABP/OTAA join attack, network/session key reuse, frame counter replay, downlink injection on TTN/Helium-style networks, sub-GHz protocol replay (KeeLoq garage doors, fixed-code remotes, TPMS spoofing, smart plug telemetry), HackRF / RTL-SDR / Flipper Zero workflows, signal analysis with Inspectrum / Universal Radio Hacker, and reconstruction of proprietary packet formats. Use for LoRaWAN deployments (smart cities, asset tracking, industrial telemetry), or any wireless device using the unlicensed 433/868/915 MHz bands (garage openers, doorbells, IoT sensors, RC equipment).
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。
IoT and embedded device security testing methodology. Covers hardware reconnaissance (UART, JTAG, SWD, SPI flash, I2C EEPROM, eMMC chip-off), firmware acquisition (vendor portals, OTA capture, flash dump, binwalk extraction), firmware analysis (filesystem mounting, binary triage, hardcoded secrets, default credential discovery), bootloader attacks (U-Boot console, secure-boot bypass, fault injection), runtime attacks on embedded Linux/RTOS (busybox CVEs, MTD writes, /dev/mem), wireless protocol attacks (Zigbee, BLE, Z-Wave, LoRaWAN, Thread/Matter, sub-GHz), MQTT/CoAP/Modbus/BACnet/OPC-UA exploitation, mobile companion app analysis, cloud-IoT API abuse, and side-channel/glitching basics. Use for IoT pentest, smart-home assessment, ICS/OT testing, or embedded vulnerability research.
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。
KRACK (CVE-2017-13077..082) and FragAttacks (CVE-2020-24586..588 + 26139-26147) — key reinstallation, fragmentation, and aggregation attacks against WPA2 supplicants. Covers Vanhoef's test scripts, viability against modern patched stacks (mostly mitigated post-2021), residual unpatched embedded devices and IoT vendors, and the practical limitations of these attacks in modern engagements. Use when assessing legacy supplicants, embedded clients, or vendors with poor patch cadence.
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。
Z-Wave attack methodology — sniffing with Z-Force / EZ-Wave / RTL-SDR + ZniffMobile, S0 (legacy) network-key derivation flaw and key reuse, S2 (modern) ECDH commissioning analysis, replay/injection on unauthenticated nodes, default-key brute-force on test deployments, and home-automation hub pivots. Use when targeting Z-Wave smart home devices (door locks, sensors, garage controllers) — common in mid-2010s smart home deployments still in production.
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。
Start or patch Dynamo router modes and run router endpoint smoke checks. Use for round-robin, KV-aware, least-loaded, or device-aware routing setup; use recipe-runner for recipe deployment and troubleshoot for failure diagnosis.
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。
Scan any agent skill for security risks before you install or use it. Powered by Tencent Zhuque Lab A.I.G (AI-Infra-Guard). 100% local static analysis — no file contents or credentials leave your device. Compatible with CodeBuddy, Cursor, Windsurf, Claude Code, OpenClaw and more. Triggers on: `这个 skill 安全吗`, `skill 安全扫描`, `检查 skill 安全`, `audit skill`, `scan skill`, `check skill safety`, `analyze skill`, `inspect skill`, `verify skill`, `skill security`, `skill supply chain`. Do NOT trigger for general agent usage, full system health checks, project debugging, or normal development.
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。
Draws flat vector spot illustrations of tall, long-legged people doing design and coding things, with solid black as a main colour. Traits: no outlines on coloured fills; black as big solid shapes (hair, trousers, shoes, devices) and as thin detail lines; denim blue, leaf green, sun yellow and tomato red on cream skin; dot eyes, round glasses and blush; big shoes on one thin black ground line; two stars and two plus-sparkles as the only decor. Covers palette, line weights, characters, heads, hands, props, composition and a parts helper for limbs, shoes, heads, hands and paper props. Use when someone asks for flat illustration with black, flat people or character illustration, friendly product or SaaS illustration, or art for empty states, onboarding, 404 and error pages, blog headers, feature spots, marketing cards or app screens in this style.
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。
Draws original design-and-coding scenes as SVG cards (480x360) in a framed-panel cartoon style. Identifying traits are a soft pastel card with a periwinkle-violet panel framed in thin navy, outline-only clouds and a pale lavender floor band with dashes inside it; L-shaped corner brackets outside two corners; white speech bubbles, UI cards and bushes that break out over the frame edge; and a semi-real character with a thin navy outline (1.25 px), muted flat fills, navy hair with violet highlights, blush and short paired hatch marks. Covers people working with laptops, screens, UI parts and devices. Use when someone asks for a framed panel, window-frame or picture-frame illustration, pastel violet cartoon, outlined flat character scene, empty-state, onboarding or 404 art, a blog header, a feature spot, a support, about or team page illustration, or a marketing card with a person.
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。
Draws original spot illustrations of one hero object as SVG cards (480x360) in a teal-on-white style. Identifying traits are a single object in a fixed oblique 3/4 view showing top, front and right faces; white main faces with teal only on side faces and accents; thin dark-teal outlines on every edge; a hard, unblurred deep-teal drop shadow pushed down-right; and two to four small floating extras (coins, tilted UI cards, badges) with their own hard shadows, plus 4-point sparkles. No people. Covers design and coding concepts as objects: keyboards, databases, git graphs, inboxes, devices. Use when someone asks for a teal spot, teal icon illustration, isometric spot object, 3D line spot, empty-state art, onboarding step art, a 404 or error spot, a feature spot, a blog header object, a marketing card or a product icon scene in teal, mint or green-teal.
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。
Client-side request-signing and anti-bot token reversal for bug bounty — when a request carries a sign/sig/hmac/token/nonce/timestamp/X-Sensor header that Burp Repeater cannot replay, recover the signer just enough to reproduce the request outside the client. Packet-first staging (capture real request → prove replay works → only reverse if replay fails) across the locate→recover→runtime→validation→replay spine. Covers tracing backward from the signature field (writer→builder→entry→source), isolating user-mutable sign inputs (timestamp/nonce/deviceId/body) vs constants (secret key), hooking fetch/XHR in DevTools, JS deobfuscation basics (webpack/wasm/JSVMP), and the bounty payoff: reach the protected API to then hunt IDOR/auth/business-logic. Use when Burp/mitmproxy replay of a signed or anti-bot-gated request fails and you suspect a client-computed field is blocking you.
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。
Hunt Session Management vulnerabilities — session fixation (no regeneration on login), insufficient invalidation on logout / password-change / email-change, predictable or low-entropy session IDs, JWT-as-session with no exp/revocation, refresh-token rotation/reuse-detection gaps, OAuth/SSO session linkage, device-bound-session (DBSC) downgrade, and cookie attribute issues (Secure/HttpOnly/SameSite/__Host-). Validate with TWO real sessions (attacker A + victim B), body-diff every 200, and OOB confirmation for theft chains. Medium to Critical (fixation→admin hijack, no-invalidation→persistent ATO).
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。
Return current frame timing, FPS, vsync, target frame rate, threading mode, and graphics device type from Unity Time / QualitySettings / SystemInfo.
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。
Opt-in evidence-first causal diagnosis for bugs, browser or app/device failures, flaky behavior, and performance regressions. Activate only when the user explicitly invokes `$diagnose` or explicitly asks to use the named diagnose skill. Do not activate merely because the user mentions a bug, asks why something failed, requests debugging or a fix, or describes unexpected behavior. Once explicitly invoked, locate likely causes, instrument relevant boundaries with extensive structured logging, reproduce the issue, analyze the collected logs, rank causes with confidence scores, and pursue 100% operational confidence while probes can increase confidence.
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。
Inspect and profile React Native component trees from agent-device. Use for React Native performance, profiling, props, state, hooks, render causes, slow components, excessive rerenders, or questions like why a component rerendered.
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。
Query openFDA API for drugs, devices, adverse events, recalls, regulatory submissions (510k, PMA), substance identification (UNII), for FDA regulatory data analysis and safety research.
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。
Official NVIDIA-authored guidance for PhysicsNeMo ShardTensor domain parallelism — integrate domain parallelism into training/inference scripts (new or existing) with DDP or FSDP2, write and register shard patches to enable new layers/ops, and bootstrap multi-GPU correctness tests. Use when working with ShardTensor, scatter_tensor, domain parallelism, sequence/spatial sharding, ring attention, DeviceMesh + DDP/FSDP2 hybrid parallelism, or physicsnemo.domain_parallel. Do NOT use for generic PyTorch DDP/FSDP setup without domain parallelism, picking a PhysicsNeMo model or example (use physicsnemo-discover), or non-distributed training questions.
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 controlling NVIDIA Broadcast through MCP to apply effects, process local media, select devices, or change camera resolution, or when Broadcast is missing or too old to expose the gateway and the user wants it installed; not for Broadcast app settings outside the MCP gateway.
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。
Validate and use CPU offloading in Megatron Bridge, including layer-level activation offloading and fractional optimizer state offloading with HybridDeviceOptimizer.
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 installing, repairing, reusing, inspecting, or verifying readiness of the native NVIDIA Video Codec SDK or PyNvVideoCodec on Jetson, including the one-frame encode/decode smoke test with official samples, and when interpreting what those readiness results, including CPU-buffer and device-memory sample modes, do and do not establish.
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。
How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.
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 this skill when the user is doing hands-on NVMe-over-Fabrics storage-target work on a BlueField DPU or ConnectX NIC with DOCA STA — standing up a doca_sta DOCA Core context that accelerates the target-side NVMe-oF data path over RDMA, defining doca_sta_subsystem targets (NQN + namespaces) backed by local NVMe-PCI backend disks (doca_sta_be), checking device support via doca_sta_cap_is_supported, sizing the per-connection I/O queues, or debugging DOCA_ERROR_* from a STA call. Trigger even when the user does not say "DOCA STA" — typical implicit phrasings include "my NVMe-oF Connect never completes", "Identify Controller times out over RoCE", "16 I/O queues at depth 1024 — does this BlueField support that", "offload the nvmf target onto the DPU", or "DOCA_ERROR_IO_FAILED on an NVMe read". Refuse and route elsewhere for DOCA install, raw RDMA data movement, raw packet I/O, flow-rule programming, or initiator-side / host NVMe stack work — those belong to other skills.
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。