发现 Skills

Leverage shell_agent for autonomous file creation and modification with automatic tool selection and error recovery

Train or fine-tune language and vision models using TRL (Transformer Reinforcement Learning) or Unsloth with Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, model selection/leaderboards and model persistence. Use for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.

Diffusers Pipeline inference guide for Huawei Ascend NPU. Covers environment pre-checks, general Pipeline inference (image and video models), memory optimizations (CPU offload, attention slicing, VAE slicing), LoRA loading and merging, multi‑card inference, and locating the Diffusers API by version. Use this skill whenever a user mentions running Diffusers inference such as FLUX, SDXL, Wan, CogVideoX on Ascend NPU or needs to troubleshoot inference parameters, memory optimization, LoRA integration, or Pipeline API paths.

Muxi is a deep learning training acceleration tool that supports elastic training, automatic fault recovery, dynamic resource scheduling and other enterprise-grade distributed training capabilities.

GENErator DNA sequence generation model Ascend NPU migration Skill for migrating a HuggingFace Transformers–based Causal LM from CUDA to Huawei Ascend NPU. Covers environment setup, dependency installation, code adaptation, multi-process handling, and sequence recovery verification.

React and Next.js performance optimization guidelines. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers when working in *.tsx files for websites or web applications involving React components, Next.js pages, web pages, data fetching, bundle optimization, or performance improvements.

Deploy the doratech.cn website to a production server via SSH. Use when the user asks to deploy, publish, update, or push the website to production, or mentions "dora-website", "doratech.cn deployment", or phrases like "go live"/"deploy website".

Analyzes historical attack traces to provide intelligent recommendations for attack selection, transform effectiveness, and strategic adaptation

Performs AI-powered code quality analysis on custom WordPress code (active theme + custom plugins) using a two-pass tiered approach - quick pattern scan for known anti-patterns, followed by deep AI contextual analysis on flagged files.

Use when you suspect API keys, tokens, or passwords are hardcoded in source code or committed to git history — scans and guides safe removal without breaking existing integrations.

typing.

Single source of truth and librarian for ALL Gemini CLI documentation. Manages local documentation storage, scraping, discovery, and resolution. Use when finding, locating, searching, or resolving Gemini CLI documentation; discovering docs by keywords, category, tags, or natural language queries; scraping from llms.txt; managing index metadata (keywords, tags, aliases); or rebuilding index from filesystem. Run scripts to scrape, find, and resolve documentation. Handles doc_id resolution, keyword search, natural language queries, category/tag filtering, alias resolution, llms.txt parsing, markdown subsection extraction for internal use, hash-based drift detection, and comprehensive index maintenance.

Check Antigravity account quotas for Claude and Gemini models. Shows remaining quota and reset times with ban detection.

Use Danube's 100+ API tools (Gmail, GitHub, Notion, etc.) through MCP. Search for tools, check authentication, execute with parameters, and handle errors gracefully.

Interact with Messages app - read conversations, send messages, and check for new messages using AppleScript and SQLite database access

专业安全审计专家，精通漏洞扫描、渗透测试、合规检查和安全代码审查。帮助企业识别和修复安全风险，确保系统和数据的安全性。

Testing knowledge base for PHP 8.4 projects. Provides testing pyramid, AAA pattern, naming conventions, isolation principles, DDD testing guidelines, and PHPUnit patterns.

As a data warehouse and SQL expert, provide data lineage analysis services based on static SQL code.

SDFormat/SDF robot and simulator model generation and validation. Use when creating, editing, regenerating, inspecting, or validating `.sdf` / SDFormat files, `gen_sdf()` sources, SDF models or worlds, links, joints, poses, frames, inertials, visual/collision geometry, mesh URIs, sensors, plugins, or simulator-specific metadata. Use the URDF skill for ROS robot structure; use the SRDF skill for MoveIt planning semantics.

Run accessibility testing on a URL using the axe MCP Server — analyze violations, get remediation guidance, apply fixes, and verify. Use when asked to check accessibility, run a11y audit, or fix WCAG issues.

AI-Ready 프로젝트 부트스트랩. 사용자가 "새 프로젝트 셋업", "AI-ready 만들어줘", "프로젝트 초기화" 등을 말하면 무조건 사용. 12개 파일 + 디렉터리 결정론적 생성.

Use when executing implementation plans with independent tasks in the current session

Reviews Rust microservices in the depot/ directory with focus on Axum web framework, SQLx database patterns, Docker deployment, and WebSocket communication. Use when reviewing depot service changes (discovery, dispatch, map-api, gps-status, mapper), adding new endpoints, modifying database schemas, implementing WebSocket protocols, or debugging service integration issues. Covers Tokio async patterns, REST/WebSocket API design, PostgreSQL migrations, error handling, health checks, and Docker multi-stage builds.

Use when writing instructions that guide Claude behavior - skills, CLAUDE.md files, agent prompts, system prompts. Covers token efficiency, compliance techniques, and discovery optimization.