'fuse' 검색 결과

Provides guidance for writing and benchmarking optimized CUDA kernels for NVIDIA GPUs (H100, A100, T4) targeting HuggingFace diffusers and transformers libraries. Kernels must be kernel-builder/ABI3-compliant: no pybind11, no setup.py, TORCH_LIBRARY_EXPAND bindings only. Supports models like LTX-Video, Stable Diffusion, LLaMA, Mistral, and Qwen. Includes integration with HuggingFace Kernels Hub (get_kernel) for loading pre-compiled kernels. Includes benchmarking scripts to compare kernel performance against baseline implementations.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Weekly, heavy browser lane, read only toward the world with no exception. Works the surfaces on its own watchlist through one contiguous browser phase, captures one dated observation per real change with a verbatim quote and the exact URL, refuses to report any change against a surface it has no baseline for, and writes one capped weekly page. It types nothing that changes anything, takes no action on LinkedIn ever, sends only where you released the channel, spends only where you released it, and never touches a credential.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Monthly, on the last weekday, file work only, no browser at all. Folds the decision ledger, scores this Employee's own judgement against the metric files that measured it rather than against its own predictions, appends one outcome line per closed decision, retires a move proposed three times and never accepted, and rewrites the priorities where the evidence disagrees with them. It refuses to publish a calibration figure on too few closed decisions, changes nothing outside its own folder, proposes nothing to anybody, sends only where you released the channel, spends only where you released it, and never touches a credential.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Weekly, conditional browser lane, read only everywhere. Scores the fleet and the business for the week with a source beside every single number and no number anywhere without one, cites the Employee that owns a metric rather than recomputing it, refuses a rate below the floor, and writes one capped weekly page. It opens a browser only where the metric map names a live screen, never writes into another Employee's folder, sends only where you released the channel, spends only where you released it, and never touches a credential.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Weekdays, browser only when a source refuses to be fetched. Works the single card the standup marked next: reads the shared publishing standard and the card's specification, pulls the live result set for the primary keyword, reads the pages currently ranking, writes the body, the metadata, the internal links and the outbound authority links, generates and compresses a hero, and leaves a complete draft folder with one ready line. It never opens a publishing surface, never touches a live property, and never touches a credential.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

One full-bleed launch hero in two inks from the project's DESIGN.md: the base color and exactly one brand color, as a slowly folding ordered-dither (Bayer) field — a print/riso look, not a generic gradient. Two flat inks never blend, so there is no mud. The copy sits on a hard-edged base-ink knockout, so headline and sub copy both clear 4.5:1 by construction, and a verify script measures it across the whole timeline. Reads Core v2 tokens (and Google-compatible frontmatter), refuses to invent a palette, ships a CSS halftone poster fallback, respects prefers-reduced-motion, ~7 KiB gzip of library JS. Use for a launch page, product hero, waitlist, event or rebrand first screen that needs motion in brand color. Triggers: '셰이더 히어로', '런칭 히어로', '디더링 배경', '리소 느낌 히어로', '브랜드 컬러 배경 애니메이션', 'dither hero', 'shader hero from DESIGN.md', 'launch hero background', 'riso print hero', 'シェーダーのヒーロー', '著陸頁主視覺 shader'. Whole scroll-native landing → omd-landing / omd-aphrodite. CJK headline ink effect → omd-cjk-type-shader. Brand tokens missing → omd-init first.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

One full-bleed launch hero in two inks from the project's DESIGN.md: the base color and exactly one brand color, as a slowly folding ordered-dither (Bayer) field — a print/riso look, not a generic gradient. Two flat inks never blend, so there is no mud. The copy sits on a hard-edged base-ink knockout, so headline and sub copy both clear 4.5:1 by construction, and a verify script measures it across the whole timeline. Reads Core v2 tokens (and Google-compatible frontmatter), refuses to invent a palette, ships a CSS halftone poster fallback, respects prefers-reduced-motion, ~7 KiB gzip of library JS. Use for a launch page, product hero, waitlist, event or rebrand first screen that needs motion in brand color. Triggers: '셰이더 히어로', '런칭 히어로', '디더링 배경', '리소 느낌 히어로', '브랜드 컬러 배경 애니메이션', 'dither hero', 'shader hero from DESIGN.md', 'launch hero background', 'riso print hero', 'シェーダーのヒーロー', '著陸頁主視覺 shader'. Whole scroll-native landing → omd-landing / omd-aphrodite. CJK headline ink effect → omd-cjk-type-shader. Brand tokens missing → omd-init first.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Author and review GitHub Actions workflow YAML safely so syntactically-valid YAML can't ship a workflow that GitHub Actions refuses to run. USE FOR: editing, adding, or reviewing any file under .github/workflows/, writing run-name/name/if/env/run values that contain ${{ }} expressions, diagnosing a run that fails with 'This run likely failed because of a workflow file issue' and no jobs starting, deciding when a workflow scalar must be quoted, validating workflows with actionlint. DO NOT USE FOR: authoring application YAML unrelated to GitHub Actions, Azure Pipelines, GitLab CI, or non-workflow YAML. SCOPE: this skill covers *syntactic/structural* correctness of workflow YAML (quoting, parsing, actionlint); for *semantic and functional* workflow design (what a workflow should do, agentic-workflow behavior), see .github/agents/agentic-workflows.agent.md — the two are complementary. INVOKES: actionlint (downloaded pinned binary) plus git/grep for inspection.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Enhancement-overlay skill — the DECISION + WIRING layer for LM observability that the single-backend skills [[langsmith]], [[phoenix]], [[mlflow]] do NOT cover. Each of those installs one backend; none of them help you DECIDE which backend fits your stack/scale/budget, nor give you a one-line autolog that turns it on fast. Use when starting any LM project, before the first deploy, or the moment someone asks "why did it do that?" and there are no traces to answer with. The skill picks a backend by stack (LangSmith for LangChain/LangGraph; Phoenix for OSS/local OpenTelemetry; MLflow for ML-shops already on MLflow; Langfuse for self-host), wires one-line autolog, verifies traces land, and adds eval hooks — instrumenting BEFORE you need it. Cross-links the first-debug-move skill [[agentsop-prompt-history-inspect]]. Do NOT activate to re-teach a backend you already chose (defer to its own skill), or for non-LM ML experiment tracking with no LLM calls (that is plain MLflow).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

SkillAlchemy — One thought conceived, one goal achieved. Accept any idea or distillation target and produce an installable SKILL.md. It uses Lens to clarify the problem and LEAP to run distillation or fusion. This is the sole user-facing entry point. Use when the user asks to distill, generate a skill, fuse skills, or says, "I want to build X, but I do not know where to start."

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Lens — Add a cognitive lens to any problem. It accepts a task description and produces an enhanced description that surfaces hidden dimensions, prerequisites, and lines of inquiry—the things you do not know you do not know. Use when the user asks to brainstorm, analyze, generate a skill, distill, or fuse, or when the input is too simple and needs to be expanded.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Comprehensive backend development guide for Langfuse's Next.js 14/tRPC/Express/TypeScript monorepo. Use when creating tRPC routers, public API endpoints, BullMQ queue processors, services, or working with tRPC procedures, Next.js API routes, Prisma database access, ClickHouse analytics queries, Redis queues, OpenTelemetry instrumentation, Zod v4 validation, env.mjs configuration, tenant isolation patterns, or async patterns. Covers layered architecture (tRPC procedures → services, queue processors → services), dual database system (PostgreSQL + ClickHouse), projectId filtering for multi-tenant isolation, traceException error handling, observability patterns, and testing strategies (Jest for web, vitest for worker).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Brings task.md, changelog.md, and files.md back in sync with what the code says shipped. Reads git history and the working tree, never memory, and refuses to log features it cannot see in source. Use when a feature or fix lands, when the user says "update the docs", "sync changelog", "update files.md", "@update", or when task.md and the code disagree.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Complete UI/UX design capability — a wiki-style knowledge map of design theory (UX laws, Nielsen heuristics, usability, visual hierarchy, typography, contrast/WCAG, design systems, aesthetics-vs-conversion) fused with a hands-on execution playbook for building distinctive, production-grade frontend interfaces. Use when designing or building any UI (websites, landing pages, dashboards, components, apps), reviewing or critiquing UI/UX, diagnosing why a design doesn't convert, or styling/beautifying web interfaces. Grounds every visual decision in UX principles and avoids generic AI aesthetics.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Deliberate-choice procedure for a medium-to-high-stakes engineering fork — when the first plausible solution (the instinct, the default next-token pull) would be costly to get wrong. Fires AUTONOMOUSLY: invoke proactively whenever a fork fits, never wait to be asked — mechanical triggers include a fix failing for the 2nd-3rd time, adding a dependency, schema/migration design, deleting or deprecating things others depend on, changing a public API, choosing an architecture or stack. Refuse the premature collapse: hold real options open (urge · contrarian · synthesis · out-of-box · intuitive dots · precedent · first-principles), ground each branch in at least one fact from outside the model (codebase, docs, benchmark, spike), future-model consequences (blast radius, reversibility, maintenance, pre-mortem), collapse by deliberate choice, then try to refute the winner before acting. Log the decision and rejected branches in docs/decision_logs/. Not for routine calls — a decision worth more than one forward pass.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Builds estimates, quotes, and bids for trades and field-service businesses (HVAC, plumbing, electrical, roofing, landscaping, painting, cleaning, remodeling, general contracting, IT/AV installs). Turns a scope into line items with waste factors, burdened labor, materials, equipment, subs, permits, mobilization, contingency, overhead, profit, and tax, then writes a client-ready xlsx (live formulas) and PDF plus an internal cost sheet, with good/better/best options, add-ons, a payment schedule, exclusions, and assumptions. Use whenever someone asks for a quote, estimate, bid, proposal price, or job price, asks "how much should I charge for this job", wants to check a price or margin on a job, or is confused about markup vs margin. Computes deterministically with a script instead of mental math, because models routinely treat a 30% markup as a 30% margin and drop labor burden, waste, and mobilization.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Route Custom Diffusion workflows for concept data prep, diffusers training and SDXL, sample generation, delta tools, and CustomConcept101 evaluation.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Select appropriate Ollama models for processing sensitive but legal content. Use when: (1) llama3.2 or deepseek-r1 refuses to process content with "I cannot assist" responses, (2) Local LLM returns generic safety disclaimers instead of following instructions, (3) Need to process adult content, sex work data, or other legal-but-sensitive material through a local LLM. Also covers using /api/chat vs /api/generate for system prompt support, and structuring prompts for consistent format compliance.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Fix black camera/microphone preview in Flutter macOS apps launched via `flutter run` from a terminal emulator (cmux, iTerm, Terminal.app, VS Code integrated terminal, Cursor, etc.). Use when: (1) camera plugin reports initialized but preview is black with no frames, (2) logs show "initialization completed via timeout" or similar first-frame timeout fallback, (3) AVCaptureSession.startRunning() returns without error but zero frames arrive, (4) the app's Info.plist has NSCameraUsageDescription and entitlements (com.apple.security.device.camera, audio-input) are correctly set, (5) no permission dialog ever appears, (6) same failure mode for microphone. Root cause is macOS TCC attributing camera/mic requests to the *responsible process* (the parent terminal) rather than the Flutter app itself. TCC refuses to prompt because the terminal is hardened-runtime without a camera entitlement, and silently denies. Not a bug in the camera plugin — would reproduce with ANY camera code.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Run before pushing a PR that touched UI to find untranslated keys in any of the 21 non-English locales and to catch user-visible English strings that bypass context.l10n. Reports findings as a checklist; refuses to declare clean until all are addressed or explicitly waived. Invoke with /check-l10n.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

.NET / C# 二进制逆向。当目标是 .NET assembly（PE 头含 CLR、.exe/.dll 托管程序）、C# 编译产物（含 NativeAOT）、红队 Sharp* 工具（Rubeus / SharpHound / SharpHound 等）、.NET 混淆程序（ConfuserEx / SmartAssembly / Babel / Eazfuscator）、.NET loader / info-stealer / 套壳 malware 时使用。优先用 dnSpyEx + de4dot，需要 AI 直接操作时联动 dnSpy MCP。不用于纯 native 二进制（走 reverse-engineering / ida-reverse）。

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Cloudflare Hyperdrive for Workers-to-database connections with pooling and caching. Use for PostgreSQL/MySQL, Drizzle/Prisma, or encountering pool errors, TLS issues, connection refused.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Scaffold and amend the L1 Investment Constitution — [ALL_CAPS] placeholder bootstrap, SemVer bump rules, Sync Impact Report, MINOR/MAJOR re-examination dispatch after the gate-5 budget confirm. Thesis creation stays refused until ratification (Q83).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Create a research thesis — theses/{nnn}-{slug}/ via mkdir-as-CAS, pillar-prioritized spec.md, initialized thesis.md, and the thesis-quality checklist. Refuses creation while the workspace constitution is unratified (spec 046 Q83 A+).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.