aktonay/skillz

最近记录的来源活动: 2026年10月29日 10:13

SkillsMP 收录数据更新: 2026年11月12日 12:52

已收集 skills: 49

GitHub 星标: 0
GitHub Forks: 0

用 AI 助手安装

复制这段 Prompt，发给你正在使用的 AI 助手。

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/aktonay/skillz.

仓库 README

仓库 README 来源：README.md

查看 README 源文件

展开完整 README

# ⚡ SKILLZ — Universal AI Agent Skills Engine & Runtime Harness ``` ███████╗██╗ ██╗██╗██╗ ██╗ ███████╗ ██╔════╝██║ ██╔╝██║██║ ██║ ╚══███╔╝ ███████╗█████╔╝ ██║██║ ██║ ███╔╝ ╚════██║██╔═██╗ ██║██║ ██║ ███╔╝ ███████║██║ ██╗██║███████╗███████╗███████╗ ╚══════╝╚═╝ ╚═╝╚═╝╚══════╝╚══════╝╚══════╝ ⚡ UNIVERSAL AGENT SKILLS & RUNTIME HARNESS ⚡ ``` > **The universal skills engine, runtime harness, and one-command installer for Claude Code, Cursor, Google Antigravity, OpenAI Codex, Qwen, GLM, Windsurf, OpenHands, Cline, Roo, and all AI coding agents.** Equips your AI coding agents with an autonomous execution harness, software engineering discipline, token optimization, enterprise design systems, and 162+ production engineering skills through a single unified engine. <img src="assets/skillz_architecture.png" alt="Skillz Architecture Diagram" width="920" /> --- ## 🚀 Instant 1-Command Universal Installation Install and activate all **162 production skills** across every AI coding agent on your machine in a single command: ### 🪟 Windows (PowerShell) ```powershell irm https://raw.githubusercontent.com/aktonay/skillz/main/install.ps1 | iex ``` ### 🍎 Linux & macOS (Bash) ```bash curl -fsSL https://raw.githubusercontent.com/aktonay/skillz/main/install.sh | bash ``` ### 📦 Node / NPX ```bash npx git+https://github.com/aktonay/skillz.git install --all --target all ``` ### 🤖 For AI Coding Agents (Claude Code, Cursor, Antigravity, Codex, etc.) Simply give your AI agent this prompt: > *"Install the Skillz universal pack across all agents by running:* > `skills install --all --target all`" --- ## 🧙 Interactive 6-Question Role Setup Wizard & Persona Profiles Different engineering roles require different agent capabilities. **Designers don't need python test runners or cloud infrastructure bloat**, while **QA test engineers don't need 43 UI design systems or animation timelines**. Skillz includes an **interactive 6-question profiling wizard** that tailors your agent's skills to your exact workflow in seconds: ```bash skills wizard # Runs the interactive 6-question setup questionnaire skills profiles # Displays all available role profiles and exclusions ``` ### 📋 The 6 Profiling Questions | # | Question Focus | What It Configures | Options | |---|---|---|---| | **1** | **Primary Role & Agent Archetype** | Determines core domain specialization and filters out irrelevant domain bloat. | `1` 🎨 **Designer**: 43 UI design systems, animations, styling *(skips QA & testing)* `2` 🛡️ **QA Engineer**: Pytest, verification gates, TDD, debugging *(skips 43 UI designs)* `3` ☁️ **Backend Architect**: GCP, Docker, APIs, reliability, architecture *(skips UI design)* `4` 🧠 **AI Agent Orchestrator**: ECC Harness, Harness² OS, superpowers, caveman *(skips UI)* `5` 🚀 **Full-Stack Founder**: Curated 12 top UI designs + TDD + backend + QA + cloud `6` ⚡ **Universal Power User**: Complete 162-skill arsenal across all 11 categories | | **2** | **Engineering Rigor & Verification** | Sets the strictness of testing and completion gating. | `1` 🔒 **Strict & Contract-Driven**: Mandatory TDD, halt-on-failure contracts `2` 🔄 **Pragmatic & Agile**: Iterative plan-before-act, fast feedback loops `3` ⚡ **Rapid & Creative**: Frictionless prototyping without test barriers | | **3** | **Token Efficiency & Communication** | Configures agent verbosity and token compression. | `1` ⚡ **Caveman Token Compression**: Saves 60%+ tokens with ultra-terse execution `2` 📖 **Standard & Conversational**: Full explanations, detailed reasoning notes | | **4** | **Optional Feature Packs** | Allows adding cross-functional packs outside the primary role. | `1` Cloud & Infrastructure • `2` Python & QA Automation • `3` UI/UX & Design • `4` Cursor Tools • `5` Pure Role Focus | | **5** | **Target AI Coding Agent(s)** | Selects which agents to install and link. | `1` All detected agents on PC • `2` Claude Code • `3` Cursor • `4` Antigravity • `5` OpenClaw • `6` Windsurf | | **6** | **Filesystem Link Strategy** | Selects live synchronization or static copies. | `1` 🔗 **Zero-Drift Symlink / Junction** (Edits in repo reflect immediately) `2` 📁 **Standalone Directory Copy** (Isolated files) | ### 🎯 Role Profiles Breakdown (Clean Specialization) | Profile ID | Target Role | Skills Included | Categories Excluded (No Bloat) | Example 1-Liner | |---|---|:---:|---|---| | `designer` | 🎨 **Frontend & UI/UX Designer** | **65** | 🐍 `python-testing`, ☁️ `cloud`, 🧠 `agent-harness`, 🧭 `ide-meta` | `skills install --profile designer` | | `qa` | 🛡️ **QA & Reliability Engineer** | **32** | 🎨 `design` *(skips all 43 CSS & UI design libraries)* | `skills install --profile qa` | | `backend` | ☁️ **Backend & Cloud Architect** | **49** | 🎨 `design` *(skips all 43 CSS & UI design libraries)* | `skills install --profile backend` | | `agent-harness` | 🧠 **AI Agent Orchestrator** | **53** | 🎨 `design` *(skips all 43 CSS & UI design libraries)* | `skills install --profile agent-harness` | | `fullstack` | 🚀 **Full-Stack Solo Founder** | **94** | Curated balance: top 12 UI designs + full TDD + cloud + QA | `skills install --profile fullstack` | | `all` | ⚡ **Universal Power User** | **162** | None *(All 11 categories active)* | `skills install --profile all` | --- ## 🎯 What Skillz Does Skillz upgrades autonomous coding agents from simple chat assistants into disciplined, high-velocity engineering partners. It provides: ### 1. 🛡️ Autonomous Agent Harness OS & Verification Loops - **Rigid Sequential Engineering Pipeline**: Enforces a strict multi-step workflow (`Plan` ➔ `Test` ➔ `Implement` ➔ `Review` ➔ `Verify` ➔ `Remember`). - **Gated Progress & Halt-on-Failure**: Forbids marking tasks complete without deterministic test passes. No hallucinated victories or false reports. - **Recursive Agent Harnessing**: Dynamically adapts agent tools, verification probes, and context windows for complex open-world tasks. - **Contract & Invariant Enforcement**: Locks task scopes so agents cannot make unauthorized modifications outside declared requirements. ### 2. ⚡ Token & Latency Compression Engine - **60%+ Token Reduction**: Cuts verbose pleasantries, conversational filler, and redundant boilerplate from model generation loops. - **Accelerated Turnaround**: Slashes round-trip response times and token costs while keeping code diffs, architectural decisions, and diagnostics razor-sharp. - **Adaptive Terse Mode**: Dynamically compresses agent reasoning and memory logs during long-running tasks. ### 3. 🧪 Superpowers Engineering Methodology & TDD - **Test-Driven Development (TDD) Invariants**: Forces agents to write failing unit/integration tests before writing implementation code. - **Git Worktree Isolation**: Runs complex refactors, experiments, and subagents in throwaway git worktrees to keep your working branch clean. - **Autonomous Subagent Audits**: Spawns isolated peer-review passes to audit code quality, edge cases, and security vulnerabilities prior to commits. - **Bug Reproduction Gating**: Requires creating reproducible test cases before any bugfix is accepted. ### 4. 🎨 Enterprise UI/UX & High-Performance Design Systems - **79 Curated Design Systems**: Bento grid, sleek SaaS dark modes, glassmorphism, neo-brutalism, minimal luxury, and accessible high-contrast themes. - **192 Production Color Palettes & 74 Font Pairings**: Curated HSL tokens and Google Font pairings calibrated for enterprise aesthetics. - **70+ Brand Design Specifications**: Production-ready styling guides modeled after Linear, Stripe, Apple, Vercel, and modern tech leaders. - **Fluid Animation & Micro-Interactions**: GSAP timeline controllers, animated React components, interactive WebGL shader canvases, and 3D scroll scrubbers. - **Anti-AI-Slop Visual Quality Gates**: Automated inspection checks that eliminate generic AI-generated interfaces, ensuring proper hierarchy, padding rhythms, contrast ratios, and interactive states. ### 5. ☁️ Cloud, Systems & Reliability - **Production Cloud Patterns**: Google Cloud architecture, Cloud Run deployments, Docker multi-stage containers, Kubernetes manifests, and Terraform IaC. - **Gravitas Invariant Checking**: Deterministic checks that validate contracts and verify mutations with real test evidence. - **Continuous Registry Synchronization**: Validates frontmatter, licenses, and dependencies across all skills. ### 6. 🔄 Universal Cross-Agent Symlinking & Zero Disk Waste - **Directory Junctions & Symlinks**: All 20+ agent environments link back to a single canonical source on your machine. - **Instant Synchronization**: Modifying any skill in Skillz updates Claude Code, Cursor, Antigravity, and all other agents in real time without reinstallation. - **Smart Deduplication & Skip Engine**: Checks existing links and automatically skips up-to-date skills while refreshing stale links to the latest revision. --- ## 🤖 Supported Agents Matrix (20+ Targets) | Agent | Config Directory | Integration Protocol | |---|---|---| | **Claude Code** | `~/.claude/skills` | Standard `SKILL.md` frontmatter + CLI | | **Cursor** | `~/.cursor/skills-cursor` | Agent tools, prompt rules & MDC rules | | **Google Antigravity & Gemini CLI** | `~/.gemini/config/skills` | Native Customization System & Sidecars | | **OpenAI Codex CLI** | `~/.codex/skills` | Instructions & Tool Schemas | | **Qwen Agent** | `~/.qwen/skills` | Skill runners & execution loops | | **GLM / ZCode / ZEncoder** | `~/.zcode/skills` | Standard agent tools & templates | | **Windsurf (Codeium)** | `~/.codeium/windsurf/skills` | Cascade rules & prompt hooks | | **OpenHands** | `~/.openhands/skills` | Tool wrappers & environment runners | | **Cline** | `~/.cline/skills` | Custom instructions & skill definitions | | **Roo Code** | `~/.roo/skills` | Mode instructions & custom skills | | **Trae** | `~/.trae/skills` | Agent skill integration | | **GitHub Copilot** | `~/.copilot/skills` | CLI extensions & agent rules | | **Continue.dev** | `~/.continue/skills` | System prompts & custom tools | | **Kiro / Adal / Pochi / OpenClaw** | Respective agent directories | Standard multi-agent skill layout | --- ## 📦 162 Production Skills Breakdown | Category | Skills | What They Do | |---|:---:|---| | **Agent Harness OS** | 11 | Multi-stage execution pipelines, test gates, invariant checkers, recursive harnesses | | **Superpowers** | 12 | Test-Driven Development, git worktrees, systematic debugging, subagent peer reviews | | **Token Efficiency** | 8 | Terse mode, token compression, prompt optimization, latency reduction | | **UI/UX & Design** | 43 | Design tokens, color palettes, motion choreography, bento grids, anti-AI-slop gates | | **Development & Review** | 18 | Code review, performance auditing, security scanning, refactoring workflows | | **Google Cloud & Infra** | 13 | Cloud Run, Kubernetes, Terraform modules, Docker deployment setups | | **Cursor Tools** | 21 | Rule generators, context builders, multi-file editors, scratchpad workflows | | **Reliability Engines** | 5 | Gravitas contract verification, Snapdec evaluation and scoring | | **Python Testing** | 9 | Pytest runners, coverage gates, pre-commit fixtures, linting suites | | **Engineering Stacks** | 2 | Full-stack engineering workflows and dependency graph analyzers | | **IDE Customization** | 3 | Editor extension guides, sidecar configs, custom tool setups | *For the complete catalog of individual skills, see [CATALOG.md](CATALOG.md).* --- ## 🛠️ CLI Command Reference The global `skills` and `skillz` commands manage your entire skill catalog: ```bash # Interactive 6-Question Setup Wizard skills wizard # View all Persona Profiles and Category Exclusions skills profiles # Install by Persona Profile (Skips irrelevant domain bloat) skills install --profile designer --target all # 65 UI design & animation skills (skips QA & testing) skills install --profile qa --target all # 32 QA & pytest skills (skips 43 UI designs) skills install --profile backend --target all # 49 cloud & systems skills (skips 43 UI designs) skills install --profile fullstack --target all # 94 curated founder essentials # Universal 1-Click Installation (Installs all 162 skills across all categories) skills install --all --target all # Update and re-sync all latest skills skills update # Install a specific skill to a specific agent skills install ecc --target claude skills install ui-ux-pro-max --target cursor # Install all skills in a category skills install --category design --target all # Show installation status and linked agents on your PC skills status # Search across all 162 skills by keyword skills search "animation" # View complete metadata, instructions, and preview of a skill skills info harness2 # Lint and validate all SKILL.md files skills validate # Rebuild machine-readable registry and catalog skills sync # Scaffold a new custom skill skills new my-custom-skill agent-harness # Open a skill in your editor skills edit caveman # Launch the interactive local visual dashboard skills ui ``` --- ## ⚙️ How It Works Under the Hood ### 1. Smart Deduplication & Skip Logic When you run `skills install`, Skillz inspects the destination path: - If a skill is already linked to the latest canonical source, it outputs `⏩ Skipped (already latest)` and moves to the next item instantly. - If a link points to a stale revision or an old copy, it replaces it with the latest canonical junction. - If multiple skills or agents share paths (such as Gemini & Antigravity), target paths are automatically deduplicated. ### 2. Standard YAML Frontmatter Every skill follows the universal agent standard in its `SKILL.md`: ```yaml --- name: skill-name description: Concise summary of what this skill does and when to activate it license: MIT --- ``` Agents natively parse this frontmatter to auto-discover when to invoke the skill. --- ## 📄 License MIT License. Free for developers, teams, and autonomous AI agents.

正在加载仓库 README…

这个仓库中的 skills

职业分类待补全

skill

职业分类

描述

更新

ecc

未分类

Everything Claude Code (ECC) / Agent Harness OS. Imposes a disciplined, multi-step engineering pipeline (plan, test, implement, review, verify, remember) with strict verification gates, memory persistence, and anti-slop rules.

原文语言：英语

2026-10-29

harness2

未分类

Harness²: Recursive Agent Harnessing (Google Research / arXiv:2610.14224). Dynamically adapts and self-improves the agent execution harness on-the-fly using dual-level recursion (Propose-Probe-Compose and Harnessing Recursion).

原文语言：英语

2026-10-29

canvas

未分类

A Cursor Canvas is a live React app that the user can open beside the chat. You MUST use a canvas when the agent produces a standalone analytical artifact — quantitative analyses, billing investigations, security audits, architecture reviews, data-heavy…

原文语言：英语

2026-10-29

env-setup

未分类

Explain, inspect, configure, and troubleshoot Cloud Agent development environments. Use when the user asks about environment setup, changing/improving the environment, or triggering/testing a build.

原文语言：英语

2026-10-29

migrate-to-builds

未分类

Test that a Cloud Agent environment will work with prebuilt environment builds and recommend any required changes. Use when the user wants to migrate to builds, test build compatibility, or follow the Builds page setup-agent flow.

原文语言：英语

2026-10-29

new-repo

未分类

Create a Cursor-hosted repo for the current project and push it. Use when a project has no git remote yet and the user asks to create a repo on Cursor, push the project to Cursor, or set up a Cursor-hosted remote; installs and signs in the origin CLI when…

原文语言：英语

2026-10-29

origin

未分类

Install, sign in, update, or repair the origin CLI for repos hosted on Cursor (origin.cursor.com). Use when the CLI is missing or outdated, git push/pull fails to authenticate against origin.cursor.com, or the user asks to set up, clone, or configure…

原文语言：英语

2026-10-29

rename-chat

未分类

Rename the current chat to match its focus. Use only when the user invokes /rename-chat. Optional text after the command steers the title.

原文语言：英语

2026-10-29

未分类

Save, back up, or share the current project on Cursor — creates a repo (a saved, versioned copy that isn't public) even for users who have never used git, handling all setup automatically. Use when the user asks to save, back up, publish, or share their…

原文语言：英语

2026-10-29

未分类

Wait for external events (GitHub or Origin CI/PR, Slack messages, Linear issues) by subscribing with the cursor-subscriptions MCP tools instead of polling.

原文语言：英语

2026-10-29

walkthrough-artifacts

未分类

Create walkthrough artifacts (screenshots and screen recordings) that prove code changes work. Use when finishing tested changes and uploading demo evidence for the user.

原文语言：英语

2026-10-29

awesome-design-md

未分类

Apply production-grade brand design systems from 70+ top companies (Apple, Stripe, Linear, Vercel, Airbnb, Ferrari, Figma, Coinbase, etc.) into any project using portable DESIGN.md files.

原文语言：英语

2026-10-29

canvas-ui

未分类

Live WebGL/WebGPU shaders running over interactive HTML DOM elements via HTML-in-Canvas. Text remains selectable, links and forms stay fully clickable while being distorted or refracted.

原文语言：英语

2026-10-29

dimensional-drawing

未分类

Make an Apple-style monochrome dimensional drawing of a UI component or physical product — title-block summary plus one dense drawing sheet with variable-based dimensions. Use when asked for an anatomy, spec sheet, redline or dimensional drawing of a…

原文语言：英语

2026-10-29

emilkowalski-skills

未分类

Master skill bundle by Emil Kowalski (Vercel/Linear design engineer) for motion design, micro-interactions, animation audits, spring physics, and Apple-grade UI polish.

原文语言：英语

2026-10-29

google-design-md

未分类

Standardized format and tools for describing design systems in single machine-readable, human-friendly DESIGN.md files combining YAML frontmatter design tokens and markdown prose.

原文语言：英语

2026-10-29

hallmark

未分类

Design skill for websites and new pages that look designed, not generated. Use when the user asks to build or design something new, or names Hallmark. Never use it on an existing design unless the user names Hallmark, as in `/hallmark redesign <path>`.

原文语言：英语

2026-10-29