機器學習

Migrate pyproject.toml from pre-PDM 2.0 syntax to modern PEP-compliant format. Focuses on dev-dependencies to dependency-groups conversion and PEP 621 project metadata. Integrates with Context7 for latest PDM documentation.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics.

Use PROACTIVELY when setting up end-to-end testing, debugging UI issues, creating visual regression suites, or automating browser testing. Uses Playwright with LLM-powered visual analysis, screenshot capture, and fix recommendations. Zero-setup for React, Next.js, Vue, Node.js, and static sites. Not for unit testing, API-only testing, or mobile native apps.

HTML 템플릿과 CSS 스타일 수정 가이드입니다. 템플릿 스타일 변경, 다크모드 수정, CSS 작업이 필요할 때 사용합니다.

Create cutting-edge WordPress plugin admin interfaces using 2026 design trends for AI-powered SEO tools. Use this skill when building WordPress admin dashboards, settings pages, or React-based plugin UIs. Implements fluid organic layouts, glassmorphism effects, Transformative Teal color theory, micro-delight interactions, and spatial depth. Optimized for AI-first experiences with humanized design elements that avoid generic AI aesthetics. Generates production-ready WordPress admin code following 2026 frontend best practices.

Эксперт по data annotation. Используй для ML labeling, annotation workflows и quality control.

Guidelines and best practices for creating draw.io diagrams with proper formatting, font handling, and layout rules. Use when creating or editing .drawio files, generating architecture diagrams, or working with draw.io XML format. Ensures high-quality PNG output with correct Japanese text rendering and professional appearance.

Annotate passing code with protection levels (@protected/@immutable/@maintainable). Use when stabilizing working features to prevent regressions.

Data validation and pipeline testing utilities for ML training projects. Validates datasets, model checkpoints, training pipelines, and dependencies. Use when validating training data, checking model outputs, testing ML pipelines, verifying dependencies, debugging training failures, or ensuring data quality before training.

Docker and Docker Compose reference for container deployment, networking, volumes,and orchestration. Includes Proxmox hosting and LXC comparison patterns.Use when working with docker-compose.yaml, Dockerfiles, troubleshooting containers,or planning container architecture.Triggers: docker, compose, container, dockerfile, volume, network, service, lxc.

Two-phase workflow that generates explicit YAML plan before execution, with approval gates and checkpointing. Use for complex multi-file changes, refactoring, bug fixes with dependencies, or any task where mistakes are costly. Achieves 90% first-pass success rate vs ad-hoc changes. Provides audit trail, rollback capability, and explicit approval points. Triggers on "refactor", "multi-step", "plan then execute", "complex changes", "safe refactoring".

Draft GitHub issues as YAML files from plans or requirements. Use this skill when:- User says "draft issues" or "create issue draft", or suggests refinements to the draft issues- Converting plans/requirements into GitHub issue format- User provides a plan file path (docs/plans/*.md)- Need to structure multiple related issues with parent-child relationshipsGenerates YAML files in tmp/issues/ for review before publishing to GitHub.

Universal data cleaning skill that analyzes, cleans, and standardizes datasets across CSV, TSV, Excel, Parquet, JSON, XML, and YAML formats. Generates Python scripts for efficient processing of large datasets (1M+ rows) without loading data into LLM context.

Build AI applications on AWS using Bedrock, SageMaker, and AI/ML services with best practices for enterprise deployment

Generate Nature-style academic HTML publications from notes, data, or drafts. Use when asked to create a manuscript, publication, paper, or formal scientific document.

Troubleshoot dev server reachability and HMR for WebF apps on real devices (Vite/Webpack): LAN IP vs localhost, 0.0.0.0/--host binding, firewall/ports, and websocket/HMR (ws://) issues. Use when the user mentions HMR, websocket, ws://, --host, 0.0.0.0, LAN IP, firewall, or a device that can load HTML but not assets/HMR.

Compact YAML format for defining parallel task specifications with scope, boundaries, and agent assignments. Use when creating task files for parallel development.

Garante que páginas HTML sigam práticas de acessibilidade WCAG 2.1 AA. Use quando criar ou modificar páginas HTML, adicionar componentes interativos, ou o usuário mencionar acessibilidade, ARIA, leitores de tela, ou contraste de cores.

Cost estimation scripts and tools for calculating GPU hours, training costs, and inference pricing across Modal, Lambda Labs, and RunPod platforms. Use when estimating ML training costs, comparing platform pricing, calculating GPU hours, budgeting for ML projects, or when user mentions cost estimation, pricing comparison, GPU budgeting, training cost analysis, or inference cost optimization.

Format conventions for markdown knowledge vaults. Use when writing notes or when format/syntax questions arise. Covers the dual-reader principle (human + AI), YAML frontmatter, wikilinks, and sources.