「20」の検索結果

Build interactive live slide decks with Vite + React 19, TanStack Router, WebGPU shaders, Framer Motion orchestration, 1920x1080, access-gated SPA. 39 components (layout, interaction, graphics, chrome, navigation), direction-aware slide transitions, presenter mode with speaker notes and BroadcastChannel sync, Playwright PNG/PDF export, YAML-driven authoring, Cloudflare/Vercel/Netlify deploy. Triggers on: slide deck, presentation, pitch deck, investor deck, product demo, conference talk, live slides, WebGPU slides, interactive presentation, presenter mode, speaker notes, overwatch deck.

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 tool outputs exceed 2000 tokens, tasks span multiple conversation turns, sub-agents need to share state, context window is bloating, plans need to persist across summarization, terminal/log output needs selective querying, or when user mentions "offload context", "dynamic context discovery", "filesystem memory", "scratch pad", "reduce context bloat", or "just-in-time context loading".

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。

Organize, extract, prune, and verify Claude Code persistent memory files. Covers MEMORY.md as a concise index, topic extraction to dedicated files, staleness detection, accuracy verification against project state, the dual 200-line / 25KB load cap, and reachability of topic files from the index. Use when MEMORY.md is approaching either load cap, after a session produces durable insights worth preserving, when a topic section has grown beyond 10-15 lines and should be extracted, when project state has changed and memory entries may be stale, or before and after rewriting the index so that no topic file loses its only pointer.

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。

Select a `delaySeconds` value when scheduling a loop wakeup via the `ScheduleWakeup` tool or the `/loop` slash command. Covers the three-tier cache-aware decision (cache-warm under 5 minutes, cache-miss 5 minutes to 1 hour, idle default 20 to 30 minutes), the 300-second anti-pattern, the [60, 3600] runtime clamp, the minute-boundary rounding quirk, and writing a telemetry-worthy `reason` field. Use when designing an autonomous loop, when a heartbeat cadence is being planned, when polling cadence is being tuned, or when post-hoc review of loop costs reveals interval mis-sizing.

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。

This skill should be used when analyzing market breadth charts, specifically the S&P 500 Breadth Index (200-Day MA based) and the US Stock Market Uptrend Stock Ratio charts. Use this skill when the user provides breadth chart images for analysis, requests market breadth assessment, positioning strategy recommendations, or wants to understand medium-term strategic and short-term tactical market outlook based on breadth indicators. All analysis and output are conducted in English.

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 writing, reviewing, or refactoring Rust code. Routes the task to handbook chapters on idioms, lints, errors, testing, generics, and pointers under the edition 2024 and current-stable pins.

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 C++ project adopts C++20 modules: named modules, partitions, header units, the global module fragment, CMake CXX_MODULES, or a BMI lookup error. Not for flag basics: use clang.

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 C++ template error needs decoding, a template needs a concept or requires-clause instead of SFINAE, or template instantiation is slowing compilation. Not for C++20 modules: use cpp-modules.

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 working with C++20 co_await, co_yield, and co_return, implementing promise_type, sizing coroutine frames, or debugging suspended coroutines in GDB. Not for Rust async: use idiomatic-rust.

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。

Fix scattered/torn regions in hex-grid simulated annealing layouts that use "competitive flood-fill" (one tile per region per outer iteration). Use when: (1) regions of >20 tiles end up dispersed instead of forming compact blobs, (2) maxDist/sqrt(N) dispersion ratio is 10×+ worse than the ideal compact blob, (3) tiles assigned to one region show up in multiple disconnected fragments across the layout, (4) the visualization has a "trypophobia ring" pattern at the perimeter where some regions spill into a perimeter spiral, (5) an overflow spiral fallback exists in the layout code and frequently kicks in for large regions, (6) bigger regions look worse than smaller ones (small regions stay compact, big ones shred). Root cause: interleaved competitive flood-fill (round-robin one tile at a time across all regions) lets large regions get boxed in by neighbors before they finish growing — the unfilled remainder is dumped into the overflow spiral at the perimeter, far from the seed. Fix: replace with sequential region-by-region BFS — process regions in size-descending order, fully claim each region's N contiguous tiles before moving to the next.

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。

Fix Node.js CLI tools crashing inside Docker containers when host-installed node_modules require a newer Node version than the container provides. Use when: (1) "SyntaxError: Invalid regular expression flags" with /v flag in string-width or similar packages, (2) node_modules installed on host with Node 20+ but container has Node 18, (3) `npm install` with file: protocol creates symlinks that break inside Docker, (4) pnpm workspace packages become broken symlinks in containers. Covers version detection, Node binary mounting, and npm install strategies for cross-version compatibility.

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。

基于 aihot.virxact.com 最新 AI 资讯，为 AI 口播频道（试界TryWorld · 纸上算法风格）产出选题清单。由 $tryworld-koubo 在模式 B 下路由调用；被显式点名「只要选题」时也可直接触发。产出为 3-8 个候选选题（含选题名、角度、素材与原文链接、流量原理命中、优先级），可直接进入 $tryworld-paper 流程做片。

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。

试界TryWorld AI 口播视频的统一定位入口，覆盖口播全流程（选题/写稿/优化/出片）。涉及口播的请求都应触发：做口播/口播视频/口播稿/写口播稿/帮我选题/口播选题/这周做什么口播/AI 资讯盘点，或直接粘贴口播稿要求优化出片。自动识别两种模式：①直接给稿 → $tryworld-paper 优化并出片；②要选题 → $tryworld-topics 拉 AIHOT 资讯产选题清单 → 挑选 → 写稿 → 优化 → 出片。子命令：只要选题/只要写稿/只优化不出片。交付附四平台推荐发布时间。无需同时引用其他技能。

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。

Create branded 试界TryWorld AI-knowledge videos in the fixed "Paper Algorithm" (纸上算法) style with HyperFrames. Input is a Chinese voiceover script and optional images; output is a 16:9 horizontal video with Azure YunxiNeural (真·云希) voiceover, word-synced captions, ink/paper animations and unified transitions, an always-visible anti-counterfeit red seal, plus horizontal (4:3) and vertical (3:4) cover images and platform-optimized titles for Bilibili/Douyin/Xiaohongshu. Brand values are theme-driven (themes/paper-algorithm.json by default; custom brand themes supported — see references/theme-guide.md). Use when the user provides an AI 科普/教程/内容解读/知识分享 script and wants a TryWorld/试界 branded video, asks for the 纸上算法/Paper Algorithm style, wants a consistent script-to-video pipeline with covers and titles, or wants this pipeline with their own brand theme.

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。

对 DiT（扩散/多模态生成）模型执行 W8A8 动态（data-free）或 W8A8 MXFP8（data-free）量化。覆盖 HunyuanVideo、Wan2.2-T2V/I2V/TI2V、FLUX.1-dev（已迁移）、SD3、Sana、HunyuanDiT、CogViewX 等。YAML apiversion 固定 multimodal_sd_modelslim_v1，统一 MultimodalPipelineInterface 重构路径（不再保留 LegacyMultimodalPipelineInterface）。模板与字段以 msmodelslim/model/hunyuan_video/model_adapter.py（唯一已迁移到重构路径且 lab_practice YAML 已 validated 的 DiT）为参考；其它 DiT 需按各自推理仓的 `parse_args` 调整字段集合。data-free 指不引入外部激活校准数据，spec.dataset（校准集短名，短名如 wan2_2_t2v；enable_dump: false 时不参与 dump）仍必填；精度调优通过下游经验库 L2 §7 + 直接调 `quantization-expert-experience-tuning-rules/scripts/apply_rollback.py`（整层回退）扩展 exclude 列表，闭环由 `quantization-accuracy-tuning-orchestrator` 调度。

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。

量化专家经验调优库：按「L1 通用专家调优意见 → L2 结构化专家调优意见 → L3 模型专属专家调优意见」三级递进， 输出量化精度调优的完整手段（离群值抑制 / 量化方法·粒度·对称性选型 / 校准集调整 / 敏感层回退 / 模型专属策略）， 每条结论附专家原因分析与专家意见可信度等级。 本 Skill 是专家经验库，负责说明可选离群值抑制算法并询问用户选择哪些算法，以及回答「怎么调、哪些层需要回退及为什么」；不运行离群值抑制 processor，不修改 YAML、不执行量化、不做 EP 检查 / 服务化 / 任务评测。

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。

Aggregate DiT inference outputs into per-dimension VBench metrics via AISBench. Run after the quant-tuning-evaluate DiT workflow produces a directory of mp4s and the user supplies a VBench-1.0-mini JSON + pre-populated cache dir. Single mode (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。

End-to-end automated model quantization and accuracy tuning workflow. Use when user asks for automated model quantization and accuracy tuning, e.g. "自动量化", "量化调优", "一键量化", "精度调优", etc. 当用户给出多卡（≥2 卡）卡号时，量化阶段通过 EP 并行适配（msmodelslim-ep-parallel-adaptation）保证多卡 EP 并行，调优主流程仍在本 Skill。

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。

执行模型量化。通过 msmodelslim quant 依据 Practice YAML 对模型进行量化。

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。

在 W8A8 动态量化后，用 MindIE-SD 冒烟测试 DiT 量化权重是否可加载（可选 skill）。与 `quant-tuning-evaluate`（DiT 扩展节）的区别：本 skill 仅做"能跑通 + 权重可用"的轻量验证（4~16 条 prompt，几分钟出结果），不算分；后者产出视频集交 `quant-tuning-score-dit`（AISBench-VBench 16 维度）评分，是调优回路一环。不引入 vLLM-Ascend、不做精度指标评估（FID/CLIP-score 等）。

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 量化调优闭环中需要生成或修改一轮调优所需的 Practice YAML，包括敏感层分析、策略决策、写出 YAML 文件和校验。

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。

重采样：把已有多轮对话数据的每条 assistant 回答丢弃，用 verifier 模型/vLLM 服务逐轮重生成，产出可直接进训练的预分词样本（每行含 input_ids/loss_mask）。依赖支持 return_token_ids 的 OpenAI 兼容 chat 服务（服务需由用户提前就绪，本技能不负责拉起服务）。触发：用某模型对某个数据集做重采样/重生成（如“对这个数据集用这个模型做重采样”）、把对话回答用模型重写、生成预分词样本/on-policy/MTP 训练数据。

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。

专门用于 Ascend 集群 Profiling 性能数据的“快慢卡”诊断专家技能。当用户提供【集群性能数据目录/路径】并要求分析【快慢卡】、【慢节点】、【负载不均衡】或【集群瓶颈】时，必须触发此技能。该技能会自动接收集群路径，调度相关工具输出快慢卡的宏观定性与微观根因（如 Host 下发瓶颈、算子计算劣化）。

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。

Append a DiT-tuning round to history.yaml. DiT-side counterpart of the LLM-path accuracy_append.py; records practice.md5, inference_outputs, fp_baseline_outputs, and the scoring fields (scores / overall_score / loss_vs_baseline / is_satisfied) populated by quant-tuning-score-dit. Idempotent on practice_id.

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。