7,661 件 ・ 関連度順
概要と使いどころ
当用户直接要求进行数据集压缩测评，或在量化调优过程中需要使用 AISBench 评测工具输出数据集精度等场景时使用。 本 Skill 负责：使用 AISBench 数据集压缩代码、用 RBF Kernel Herding 算法生成压缩后的数据子集（coreset），并按用户目标交付两种产物之一： ① 接入量化调优主流程：产出独立的 coreset 数据集配置（不覆盖原始数据），回传「全集 config_name + coreset config_name」给编排层，供 evaluation.yaml 切换； ② 仅数据集压缩测评（不做量化调优）：基于 example 模板生成「全集测试脚本」和「子集测试脚本」两份 shell 脚本。 暂时支持 AIME 2025 和 GPQA，其他数据集需要再结合 aisbench tools目录下 herding_coreset_selector 源码进行分析。
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 Ascend NPU schedule, operator dispatch, operator launch, and Host Bound profiling issues in Ascend profiling data. Use when need to diagnose device Free time, framework/operator dispatch latency, launch latency, PYTORCH_API/CANN_API launch gaps, aclrtSynchronizeStream stalls, task queue behavior, CPU scheduling interference, GC/lock pauses, CPU affinity, or schedule-side optimization actions.
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。
Programmatic React video best practices for Remotion. Use this skill whenever the user is writing, debugging, or rendering Remotion code — compositions, useCurrentFrame animations, interpolate, spring, Sequence timing, <Img>/<Video>/<Audio> from @remotion/media, staticFile, calculateMetadata, the Studio preview, `npx remotion render`, captions, subtitles, voiceover, audio visualization, ffmpeg, transitions, light leaks, transparent video, Lottie, 3D scenes, Mapbox flyovers. Triggers especially on HtmlInCanvas / "html in canvas" requests, CRT terminal effects, glitch effects, video shaders, WebGL video, and any request to scaffold a new programmatic React video. Owns `remotion render` and everything about Remotion code. Do NOT trigger for finished MP4 ad assembly from existing slides (that is `/video-content`) or for the inspiration-to-MP4 reference-video pipeline (that is `/cmo-video`).
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。
Orchestrator skill that chains /slideshow-script → /paper-marketing → /video-content for end-to-end TikTok slideshow production. Each phase is an independent Lego block — this orchestrator is just one recipe that combines them. Produces 5 publishable TikTok videos from a single positioning angle. Triggers on "TikTok slideshow", "TikTok video", "make TikTok", "slideshow video", "TikTok 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。
End-to-end social content campaign pipeline. Takes a marketing goal and produces scheduled, on-brand social posts with visuals across postiz-supported platforms, while keeping Typefully for X/thread specialist flows and fallback coverage. Chains: CMO strategy, content writing with voice calibration, AI slop audit, selective Paper MCP visual design, and routed scheduling via postiz or Typefully. Each phase has a human gate. Use this skill whenever someone wants to plan and execute a batch of social posts — not just one-off tweets. Triggers include: "social campaign", "schedule posts", "pre-launch content", "content calendar", "social content pipeline", "build up content", "I need social media content for my launch", "help me build social presence", "plan a week of posts", "batch social content", "social media strategy", or any request involving multiple scheduled posts with optional visuals.
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。
Generate 5 different narrative scripts for visual slideshows from a single positioning angle. Each script uses a different storytelling framework (AIDA, PAS, BAB, Star-Story-Solution, Stat-Flip) producing genuinely different stories, not layout variations. Make sure to use this skill whenever the user wants slideshow scripts, TikTok content scripts, carousel copy, narrative frameworks for visual content, or says anything about writing scripts for slides or social media storytelling. Even 'write me some TikTok content' or 'I need carousel copy' should trigger this. Outputs structured YAML content specs that chain directly to /paper-marketing for visual design.
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。
Reverse-engineer any person's writing voice from their content. Paste in posts, articles, tweets, or essays and this skill launches 10 parallel Sonnet subagents to analyze every dimension of the voice, then synthesizes into a voice file. Use when someone says 'match this voice,' 'analyze this writing,' 'extract their voice,' 'make me sound like this,' 'study this person's writing,' or pastes in content they want to learn from.
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 wants to train a reusable face-faithful Soul Character identity on Higgsfield. One-time training returns a reference_id passed to Soul models in higgsfield-generate via `--soul-id`. Use this skill whenever: "train soul character", "create my Soul", "train my face", "make my digital twin", "build me an avatar", "create a character of me", "set up identity for video", "I want my face in generated images", "face-faithful character", "reusable character", "brand character training", "build a character from photos", "soul training", "soul-id create". After training, logs the soul_reference_id to `brand/assets.md`. NOT for: one-shot face swaps (use higgsfield-generate with --image), fictional/non-photo avatars (use higgsfield-generate with a prompt). Requires Higgsfield CLI + Basic plan or higher.
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。
Generate an interactive HTML brand playground that shows your brand rendered live — palette, typography, logo, voice — with a social card and OG image preview that updates as you tweak tokens. Opens in browser. The visual approval step for the /cmo flow: /visual-style writes the brand, this skill shows it. Use when the user says "show me my brand", "brand playground", "preview my brand", "what does our brand look like", "visual preview", "brand kit", "see my colors", "how does my brand look", or after /visual-style completes and the user needs to approve the visual identity before moving to content generation. Also use when someone says "I want to see it" during any brand-building conversation.
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。
Generate images using the brand's visual identity and Gemini API. Reads brand/creative-kit.md for visual style, crafts narrative prompts, and produces images via Nano Banana Pro (gemini-3-pro-image-preview). Supports on-brand and freestyle modes. Use when the user needs a blog header, social graphic, product shot, hero image, banner, thumbnail, or any generated image. Also use proactively when building content that would benefit from visuals. Triggers on "generate image", "create image", "make me an image", "blog header", "social graphic", "product shot", "hero image", "banner", "thumbnail", "I need an image", "visual for", or any request for generated artwork. Even if they just say "image" or "picture for this" — this is the 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。
Record product demos and walkthroughs for marketing assets. Two modes: quick screenshot-stitch demos via ply + ffmpeg, or polished Remotion compositions. Use when the user mentions product demo, demo video, walkthrough video, feature showcase, screen recording, GIF demo, product tour, onboarding video, visual tutorial, feature walkthrough, landing page video, hero video, product video, app preview video, or wants to show their product in action. Even if they just say 'show what it does', 'make a video of the app', 'I need a demo for my landing page', or 'record the app' — this is the skill. If someone has a working product and needs marketing assets that show it off, start here. Owns screen-recorded product demos and walkthroughs. For an AI-generated marketing or brand video with no real product footage, see /higgsfield-generate; for Remotion code and `remotion render` issues, see /remotion-best-practices; for stitching existing slides into an MP4, see /video-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。
Write copy that converts. Use when writing landing pages, emails, sales copy, headlines, CTAs, social posts, cold emails, or any persuasive text. Triggers on 'copy', 'copywriting', 'sales copy', 'landing page copy', 'cold email', 'headlines', 'write me a page', 'make this convert', 'rewrite this copy', or any request involving persuasive writing. Three modes: Generate (write from scratch), Edit (improve existing copy with Seven Sweeps), Cold Email (outbound sequences). If someone has text that needs to sell harder, this is the 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。
Turn a local talking-head / 口播 video into metadata, transcript, and an AI-composed card-based video where the agent composes cards by cloning styled templates and filling them (designing bespoke only when content needs it), then renders to MP4. Ships with the Interflow AI club sign-off ("Interflow AI 出品"). Use when the user asks for interflow-video-cut, 口播成片, talking-head repurposing, video takeaways, transcript cleanup, or turning a spoken video into a captioned card video.
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。
Instructions to make or update an app's UI so that it adapts to different Android devices including phones, tablets, foldables, laptops, desktop, TV, Auto and XR. It includes how to handle different window sizes, pointing devices (such as mouse) and text entry devices (such as keyboard) using the Compose MediaQuery API. It also covers multi-pane layouts using Navigation3 Scenes, adaptive UI components (such as buttons) with varying target sizes, and adaptive layouts (including navigation areas - nav rails and nav bars) using the Compose Grid and FlexBox APIs.
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。
Make a short product film (website hero loop, landing section, launch film, feature walkthrough, store-listing promo, social cut, illustrated story film) in which the product appears as its own real UI, drawn frame by frame from an HTML timeline, rendered with Playwright and cut with ffmpeg. Four routes: licensed footage cut against cinematic product beats, product beats only, a pointer-driven walkthrough with captions or text cards (plain or kinetic), or a cut-paper collage in which an illustrated cast tells a story over real UI captured from the code with fictional data. Use when asked for a product film, hero video, promo or demo video, launch film, explainer, walkthrough video, paper or collage film, or to recut one. Not for plain screen recordings or voice-over walkthroughs; laying a supplied music track under a finished cut is in scope.
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。
Generate HTML slide presentations (1 slide = 1 HTML file, 1280x720px) using Tailwind CSS, Font Awesome, and Google Fonts. Use when the user asks to create a new presentation deck or slide HTML files. Covers design guidelines, 43 layout patterns, component library, and PPTX conversion compatibility rules. Supports style selection (creative, elegant, modern, professional, minimalist) and theme selection (marketing, portfolio, business, technology, education).
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。
Run a premortem on any plan, launch, product, hire, strategy, or decision. Imagines it failed 6 months from now, works backward to find every reason why, then produces a revised plan. Triggers include "premortem this", "premortem my", "what could kill this", "stress test this plan", "find the blind spots", "poke holes in this", "where will this break", "am I missing anything", "what could go wrong", "future-proof this", "devil's advocate this".
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 and review autonomous or semi-autonomous agent loops — systems where an agent prompts itself, runs on a schedule or event, and iterates toward a goal without a human at the keyboard. Use this whenever the user wants to build, design, harden, or audit any kind of self-running agent, agentic workflow, control loop, background worker, cron/heartbeat/webhook-driven agent, "Ralph loop", auto-PR-fixer, monitoring agent, or any long-running task that should keep going on its own until done. Trigger this even when the user says "agentic workflow", "automation that runs itself", "an agent that keeps trying until X", or "make my agent stop running away / burning tokens" without using the word "loop". Also use it to review an existing loop's stopping conditions, guardrails, verification, and escalation paths.
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。
Software Factory 2026秋の開発フローをこのsessionに適用する。 着手前に不明点をユーザーに質問、変更が一段落したらcodexでbug確認、draft PR作成、対話コンテキストexport、インラインレビューコメントの草稿、ユーザーの指示でsanity-review、ready for review後のsquash and mergeのcommit message草稿の提示、という進め方を定める。 開発の開始時にユーザーが手動で起動する。
Iteratively study and apply elegant coding patterns. Each iteration - understand the code, research what simple and elegant code looks like, apply learnings, verify with CI. Use as a standalone refactoring pass or when the user asks to make code more elegant, simple, or idiomatic.
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。
Write or rewrite technical text with the rules of ASD-STE100 Simplified Technical English so it is clear, unambiguous, and free of AI slop. Use for documentation, READMEs, runbooks, procedures, error messages, release notes, incident reports, and API guides. Also use when the user says "STE", "Simplified Technical English", "ASD-STE100", "de-slop", "make this readable", "write for non-native readers", or asks for docs that translate well. Enforces the standard's 53 rules: 20/25-word sentence limits, one word one meaning, simple tenses, active voice, condition before command.
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. Derived from Andrej Karpathy's observations. Use ALWAYS when writing, reviewing, or refactoring code. Covers assumption surfacing, simplicity, surgical edits, and goal-driven execution with verification.
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 building functions that must survive process crashes, retry automatically on failure, run on a schedule, react to events, or maintain state across infrastructure failures — e.g., webhook handlers that drop events, flaky cron jobs, background jobs that fail mid-execution, or workflows that need to resume where they left off. Covers Inngest function configuration, triggers (events, cron, invoke), step execution and memoization, idempotency, cancellation, error handling, retries, logging, and 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。
Write a feature spec or PRD from a problem statement or feature idea. Use when turning a vague idea or user request into a structured document, scoping a feature with goals and non-goals, defining success metrics and acceptance criteria, or breaking a big ask into a phased spec.
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。