🚀 SuperDuperSkills
🇬🇧 EN / 🇪🇸 ES — Bilingual documentation.
The ultimate multi-agent skills vault & governance suite: 3,900+ curated SKILL.md files (3,500+ bundled entries from 30+ verified source libraries) for Claude Code, Gemini CLI, OpenAI Codex, Cursor, OpenCode and every major AI agent — with an interactive web catalog and a project-qualification engine.
📑 Table of Contents · Tabla de Contenidos
🌐 Overview · Visión General
🇬🇧 English
SuperDuperSkills is a universal, multi-CLI AI-agent skills ecosystem. It bundles thousands of professionally maintained agent skills — reusable, self-contained instruction packages (SKILL.md) — from 30+ verified open-source libraries, deduplicates them, organizes them into 20 categories, and installs them natively into whichever AI coding tool you use.
It also ships an Interactive Project Qualification Engine: before any code is generated, an orchestrator inspects your repository (or interviews you) to qualify which skills actually apply to your project, so agents load only what they need — saving tokens and avoiding prompt bloat.
Why it exists: AI agents (Claude Code, Gemini CLI, Codex, Cursor…) each have their own native skills format. Until now, a skill written for one tool had to be rewritten for every other tool. SuperDuperSkills is a single, tool-agnostic vault that maps each skill into every harness:
| Harness / AI CLI | Local Skills Path | Protocol / Integration |
|---|
| Claude Code | ~/.claude/skills/ | Plugin Marketplace & Hooks |
| Gemini CLI / Antigravity (AGY) | ~/.gemini/antigravity-ide/ · ~/.agents/skills/ | Direct AGY Skill SDK & Auto-load |
| OpenAI Codex CLI | ~/.codex/skills/ | Codex Plugins & Hooks |
| Cursor Agent | .cursor/skills/ · ~/.cursor/skills/ | Rules for AI & Agent Skills |
| OpenCode | ~/.config/opencode/skills/ | opencode.json & NPM plugins |
| Grok Build CLI | .grok/skills/ | Grok Plugin Trust System |
| Devin CLI | ~/.devin/skills/ | Devin Plugin Marketplace |
| Kimi Code & Factory Droid | ~/.droid/skills/ | Native Skill Specification |
🇪🇸 Español
SuperDuperSkills es un ecosistema universal de skills para agentes de IA multi-CLI. Agrupa miles de skills profesionales mantenidas por la comunidad — paquetes de instrucciones reutilizables (SKILL.md) — provenientes de más de 30 librerías open-source verificadas, las deduplica, las organiza en 20 categorías y las instala de forma nativa en la herramienta de IA que uses.
Incluye además un Motor Interactivo de Cualificación de Proyectos: antes de generar código, un orquestador inspecciona tu repositorio (o te entrevista) para cualificar qué skills aplican realmente a tu proyecto, de modo que los agentes carguen solo lo necesario — ahorrando tokens y evitando inflar el prompt.
Por qué existe: cada agente de IA (Claude Code, Gemini CLI, Codex, Cursor…) tiene su propio formato nativo de skills. Hasta ahora, una skill escrita para una herramienta debía reescribirse para las demás. SuperDuperSkills es una única bóveda agnóstica que mapea cada skill a cada harness (ver tabla de compatibilidad arriba).
✨ Features · Características
🇬🇧 English
- 🧠 One vault, every agent — Universal skill format installed natively across 8+ AI CLIs and IDEs.
- 🔁 Zero duplicates — A merge-map + index pipeline deduplicates by skill name and canonical repo.
- 🎯 Project qualification engine — CLI wizard (
qualify_project.py) and interactive Web dashboard (qualifier.html) that ask what are you building? before suggesting skills.
- 📊 Interactive web catalog — Searchable, themeable (dark/light), PWA-ready UI with live category breakdowns.
- ⚡ Zero-install CLI — Run instantly with
npx superduperskills, pnpm dlx, or the native Python CLI.
- 🛡️ Governance & safety — Mandatory “Always-First” core suite (token compression, YAGNI, spec-first, memory, anti-destructive-command rules).
- 🔌 Plugin-ready — Claude Code / Cursor / OpenCode plugin manifests, shell completions, hooks.
- 📦 Python & Node dual runtime — Same feature set from
bin/superduperskills.js or scripts/superduper_cli.py.
- 🌍 Bilingual project — README, guides and site content maintained in English and Spanish.
🇪🇸 Español
- 🧠 Una bóveda, todos los agentes — Formato universal de skills instalado de forma nativa en más de 8 CLIs e IDEs de IA.
- 🔁 Cero duplicados — Pipeline de merge-map e índice que deduplica por nombre de skill y repositorio canónico.
- 🎯 Motor de cualificación — Wizard CLI (
qualify_project.py) y dashboard web interactivo (qualifier.html) que preguntan ¿qué estás construyendo? antes de sugerir skills.
- 📊 Catálogo web interactivo — UI con búsqueda, temas claro/oscuro, lista para PWA y desglose por categorías.
- ⚡ CLI sin instalación — Ejecuta al instante con
npx superduperskills, pnpm dlx o el CLI nativo de Python.
- 🛡️ Gobernanza y seguridad — Suite core “Always-First” obligatoria (compresión de tokens, YAGNI, spec-first, memoria, reglas anti-comandos destructivos).
- 🔌 Listo para plugins — Manifiestos de plugin para Claude Code / Cursor / OpenCode, completions de shell, hooks.
- 📦 Runtime dual Node/Python — Mismas funciones desde
bin/superduperskills.js o scripts/superduper_cli.py.
- 🌍 Proyecto bilingüe — README, guías y contenido del sitio mantenidos en inglés y español.
📊 Catalog Stats · Estadísticas del Catálogo
| Metric · Métrica | Value · Valor |
|---|
🗂️ Bundled SKILL.md files (bóveda) | 3,900+ |
📦 Bundled skill entries in skills/ | 3,500+ |
| 🏷️ Skill categories | 17 |
| 📚 Verified source libraries | 30+ |
| 🤖 Supported AI CLIs | 8+ |
| 🧪 Install time (typical) | < 30 seconds |
| 📄 Core documentation files | 4 main docs + bilingual guides |
Note / Nota: counts are measured live from the repository (skills/, SKILLS-INDEX.md). The interactive site shows live per-category counters that auto-generate from the index.
🗂️ Repository Structure · Estructura del Repositorio
| Ruta / Path | Descripción (EN) | Descripción (ES) |
|---|
AGENTS.md | Mandatory global guide for agents (Core Suite + qualification protocols). | Guía global obligatoria para agentes (Suite Core + protocolos de cualificación). |
README.md | This bilingual documentation. | Esta documentación bilingüe. |
LICENSE | MIT license & copyright. | Licencia MIT y copyright. |
CONTRIBUTING.md | How to add skills & contribute. | Cómo añadir skills y contribuir. |
SECURITY.md | Vulnerability reporting policy. | Política de reporte de vulnerabilidades. |
CODE_OF_CONDUCT.md | Contributor covenant (ES/EN). | Pacto de contribución (ES/EN). |
CHANGELOG.md | Version history. | Historial de versiones. |
SKILLS-INDEX.md | Deduplicated index of 3,500+ bundled skills with source repos. | Índice deduplicado de 3,500+ skills con repos de origen. |
|
🚀 Quick Start · Instalación Rápida
🇬🇧 English — Install the vault
# 1) Clone
git clone https://github.com/camilolealdev/superduperskills.git
cd superduperskills
# 2) Linux / macOS — install into your preferred AI CLI
./install.sh --target claude # Claude Code
./install.sh --target gemini # Gemini CLI / Antigravity
./install.sh --target all # every supported harness
./install.sh --dry-run # preview without writing
# 3) Windows PowerShell
.\install.ps1 -Target claude # or: -Target all
🇪🇸 Español — Instala la bóveda
# 1) Clonar
git clone https://github.com/camilolealdev/superduperskills.git
cd superduperskills
# 2) Linux / macOS — instala en tu CLI de IA preferido
./install.sh --target claude # Claude Code
./install.sh --target gemini # Gemini CLI / Antigravity
./install.sh --target all # todos los harness soportados
./install.sh --dry-run # vista previa sin escribir nada
# 3) Windows PowerShell
.\install.ps1 -Target claude # o: -Target all
🇬🇧 English — Run the CLI (zero install)
npx superduperskills # Node runtime (anywhere)
pnpm dlx superduperskills # pnpm equivalent
python scripts/superduper_cli.py ui # native Python TUI (mouse & fuzzy search)
🇪🇸 Español — Ejecuta el CLI (sin instalación)
npx superduperskills # runtime Node (en cualquier sitio)
pnpm dlx superduperskills # equivalente con pnpm
python scripts/superduper_cli.py ui # TUI nativa Python (ratón y búsqueda fuzzy)
🪟 Windows PowerShell · Terminal (Linux/macOS)
npm install -g superduperskills # provides `sds` and `superduperskills`
sds # interactive REPL
sds ui # web dashboard at http://localhost:4242
sds update # download the skills vault (~/.superduperskills)
sds completions install --shell powershell # tab completion (bash | zsh | fish | powershell)
The npm package ships only the CLI; skills come from sds update, ~/.claude/skills, or a clone of this repo · El paquete npm incluye solo el CLI; los skills vienen de sds update, ~/.claude/skills o de un clon del repo.
Requires Node ≥ 18 and Python 3 (python, py -3 or python3). Do not type the leading $ of docs snippets in PowerShell · No escribas el $ inicial en PowerShell.
🦇 Nox, the CLI mascot · Nox, la mascota del CLI
A block-art bat that follows what the CLI is doing · Un murciélago de bloques que refleja lo que hace el CLI:
| State · Estado | Eyes · Ojos | When · Cuándo |
|---|
| idle | ● ● | Menus and headers · Menús y encabezados |
| work | > < | Spinners (wings flap ^v^ -v- _v_) · Spinners (aletea) |
| happy | ^ ^ | Task or doctor succeeded · Éxito |
| error | x x | Task or doctor failed · Error |
| sleep | - - | Session closed · Sesión cerrada |
It appears in the banner (terminals ≥ 101 columns), every command header, spinners, the doctor verdict, the exit message and the web dashboard. Disable it with SDS_NO_MASCOT=1 or NO_COLOR=1 · Se desactiva con SDS_NO_MASCOT=1 o NO_COLOR=1.
💻 CLI Command Center · Centro de Comando CLI (v6.0.7 «WorldClass»)
| Command · Comando | 🇬🇧 Purpose | 🇪🇸 Propósito |
|---|
browse [category] | 📚 Browse all skills by category (18 categories, paging, --search, --json); sds → v for the interactive browser | 📚 Explora todas las skills por categoría (18 categorías, paginado, --search, --json); sds → v para el explorador interactivo |
ui | 🖱️ Interactive TUI Dashboard (Mouse support, fuzzy search, live preview) | 🖱️ Dashboard TUI interactivo (Soporte de ratón, búsqueda fuzzy, preview) |
ask "<query>" | 🤖 Local Semantic AI Assistant with RAG & streaming | 🤖 Asistente IA semántico local con RAG y streaming |
why <skill> | 🔍 Architectural justification & token breakdown | 🔍 Justificación arquitectónica y desglose de tokens |
graph | 🕸️ ASCII Dependency & synergy graph between skills | 🕸️ Grafo ASCII de dependencias y sinergias entre skills |
benchmark | ⚡ Latency & resolution throughput benchmark | ⚡ Benchmark de latencia de resolución y rendimiento |
budget <tokens> | 💰 Token budget optimizer for context windows | 💰 Optimizador de presupuesto de tokens para ventanas de contexto |
auto-branch | 🌿 Intent-based Git branch auto-generator | 🌿 Generador automático de ramas Git basado en intención |
mode <profile> | ⚙️ Switch execution profiles (fast, deep, audit, budget) | ⚙️ Conmutar perfiles de ejecución (fast, deep, audit, budget) |
scan | Deep scan of repo stack, frameworks & dependencies | Escaneo profundo del stack, frameworks y dependencias |
list | List active skills in the manifest | Listar skills activas del manifiesto |
toggle <skill> | Enable / disable a skill |
# 🖱️ Dashboard interactivo con ratón
python scripts/superduper_cli.py ui
# 🤖 Preguntarle a la IA
python scripts/superduper_cli.py ask "How to build a clean architecture API in FastAPI?"
# 🔍 Justificar por qué usar una skill
python scripts/superduper_cli.py why emil-design-eng
# 🕸️ Analizar dependencias
python scripts/superduper_cli.py graph --target tailwind-patterns
# ⚡ Ejecutar benchmark de rendimiento
python scripts/superduper_cli.py benchmark
# 💰 Ajustar skills a un presupuesto de 8,000 tokens
python scripts/superduper_cli.py budget 8000
🎛️ Agent Orchestrator · Orquestador de Agentes
🇬🇧 English
The Agent Orchestrator evaluates your project structure and performs an interview or automatic scan to determine your real goals before it allows code generation:
graph TD
A[1. Codebase Inspection + Interview] --> B[2. Load Core Always-First Suite]
B --> C[3. Interactive Selection: CLI or Web UI]
C --> D[4. Generate ACTIVE-SKILLS Manifest]
D --> E[5. Mandatory view_file SKILL.md Read]
E --> F[6. Archify 3-Commits Architecture Trigger]
F --> G[7. High-Precision Coding]
- Automatic Inspection / Socratic Interview — detects config files (
package.json, requirements.txt, go.mod, pom.xml, build.gradle, Dockerfile, .github/workflows).
- Unconditional Core Load — always injects the 27 infrastructure, quality & dynamic security skills (
caveman, ponytail, spec-kit, token-savings, harness, claude-mem, rtk, graphify, penetration-testing-with-strix, etc.) before any other rule.
- Manifest generation (
.agents/ACTIVE-SKILLS.json) — records project phase, key goals and exact local skill paths.
- Mandatory read gate (
view_file) — no AI agent writes code without first reading each qualified skill's SKILL.md.
🇪🇸 Español
El Orquestador de Agentes evalúa la estructura de tu proyecto y realiza una entrevista o escaneo automático para determinar tus objetivos reales antes de permitir la generación de código (ver diagrama de arriba):
- Inspección Automática / Entrevista Socrática — detecta archivos de configuración (
package.json, requirements.txt, go.mod, pom.xml, build.gradle, Dockerfile, .github/workflows).
- Carga Incondicional Core — inyecta siempre las 27 skills de infraestructura, calidad y seguridad dinámica (
caveman, ponytail, spec-kit, token-savings, harness, claude-mem, rtk, graphify, penetration-testing-with-strix, etc.) antes que cualquier otra regla.
- Generación del manifiesto (
.agents/ACTIVE-SKILLS.json) — registra la fase del proyecto, objetivos clave y rutas locales exactas.
- Puerta de lectura obligatoria (
view_file) — ningún agente de IA escribe código sin leer antes el SKILL.md de cada skill cualificada.
🎨 Skill Categories · Categorías (EN + ES)
SuperDuperSkills organizes its 20 categories (see docs/assets/categories-infographic.jpg) into five fundamental groups:
- Core Suite — token-compression, YAGNI architecture, persistent memory, interactive architecture diagrams, autonomous dynamic pentesting (
usestrix/strix).
- Design & UI Craft (Anti-Slop) — Emil Kowalski's design engineering (11 skills), Impeccable (23 commands), Taste framework, Apple HIG.
- Dynamic Security & Pentesting (Strix, ECC, OWASP) — autonomous AI red-teaming with verified PoC exploits (
usestrix/strix), Spring Boot Security 6 (springboot-security), secret scanning, AgentShield LLM guardrails, threat modeling.
- Backend & DevOps — Node.js, Python, Go, Java / Spring Boot, .NET, Docker, Kubernetes, Cloudflare Workers, CI/CD pipelines.
- GenAI & Testing — TDD workflows, Playwright E2E, Gemini SDK, HuggingFace, model evaluators.
🆕 What's New in 6.0.7 · Novedades 6.0.7
- 🇬🇧 All 27 core skills now ship inside
skills/ and work from the project: agentmemory, memU, understand-anything, skills-cli and jev-agent are bundled (the first three keep their upstream LICENSE; jev-agent is an original wrapper because upstream has no licence). Some need their upstream runtime installed (see each SKILL.md). Catalog: 3,572 skills.
- 🇪🇸 Las 27 skills core ya viajan dentro de
skills/ y funcionan desde el proyecto: agentmemory, memU, understand-anything, skills-cli y jev-agent están incluidas (las tres primeras con su LICENSE upstream; jev-agent es un wrapper propio porque el upstream no tiene licencia). Algunas requieren instalar su runtime upstream (ver cada SKILL.md). Catálogo: 3.572 skills.
⚡ Core Always-First Suite · Suite Core Obligatoria
[!IMPORTANT]
REGLA DE ORDEN DE EJECUCIÓN / EXECUTION-ORDER RULE: every AI agent must load this suite first:
| # | Skill | 🇬🇧 Purpose | 🇪🇸 Propósito |
|---|
| 1 | caveman | Output compression (-75% tokens, no accuracy loss) | Compresión de salida (-75% tokens) |
| 2 | ponytail | YAGNI ladder (stdlib → native deps → minimal code) | Escala YAGNI |
| 3 | spec-kit | Specification-first development | Especificación previa (Spec Kit) |
| 4 | token-savings | Keep prompt metadata lean before work starts | Confirma skills y ahorra tokens |
| 5 | harness / agent-harness | Test harnesses & continuous validation | Arneses de pruebas y validación |
| 6 | claude-mem | Persistent architectural memory between sessions | Memoria persistente entre sesiones |
| 7 | rtk | Terminal log compression (60–90%) | Compresión de logs de terminal |
| 8 | graphify | Code knowledge graph without re-reading files | Grafo de conocimiento del código |
| 9 | archify | Mandatory architecture trigger each 3 commits | Trigger de arquitectura cada 3 commits |
| 10 | skill-seekers | Active remote-skill discovery & ingestion | Ingesta y descubrimiento de skills |
| 11 | skill-vault | Frozen, versioned skill vault | Bóveda congelada y versionada |
| 12 | all-deploy | Universal deployment scripts | Despliegue universal |
| 13 | context-mode | Dynamic context-window management | Gestión dinámica del contexto |
| 14 | aprende-skill | Fast domain assimilation & synthesis | Asimilación rápida de dominios |
| 15 | agentshield | Prompt/secret sanitization, destructive-command guard | Desinfección y guardas de seguridad |
Full operating details live in AGENTS.md · Los detalles operativos completos viven en AGENTS.md.
📚 Documentation · Documentación
🛠️ Development · Desarrollo
🇬🇧 English
# Rebuild the deduplicated index after adding skills
python build_index.py
# Rebuild the minified web site (canonical: web/skills-site.html)
bash web/build.sh
# Rebuild preview / mirror copies
cp web/skills-site.html docs/index.html
cp web/skills-site.min.html docs/index.min.html
CI (.github/workflows/build-site.yml) does the same automatically on pushes that touch web/ or index.html, then commits the built artifacts. PRs touching web/** or docs/** get a live preview via .github/workflows/pr-preview.yml.
🇪🇸 Español
# Reconstruir el índice deduplicado tras añadir skills
python build_index.py
# Reconstruir el sitio web minificado (canónico: web/skills-site.html)
bash web/build.sh
# Reconstruir copias de preview / espejo
cp web/skills-site.html docs/index.html
cp web/skills-site.min.html docs/index.min.html
El CI (.github/workflows/build-site.yml) hace lo mismo automáticamente en pushes que toquen web/ o index.html, y hace commit de los artefactos construidos. Los PRs que toquen web/** o docs/** reciben un preview en vivo vía .github/workflows/pr-preview.yml.
👤 Author & Organization · Autor y Organización
🇬🇧 Built and engineered by Camilo Leal (camilolealdev) and powered by Easy Marketing. Questions, enterprise integrations, feature ideas, and skill submissions are welcome via GitHub Issues.
🇪🇸 Construido e ingeniado por Camilo Leal (camilolealdev) con el respaldo de Easy Marketing. Preguntas, integraciones empresariales, ideas y envíos de skills son bienvenidos vía GitHub Issues.
Contributors · Contribuidores
Thanks to every open-source library bundled in this vault (see skills-inventory.md) and to the global community that keeps agent skills growing. You can join them — see CONTRIBUTING.md.
⚖️ License · Licencia
MIT License
Copyright (c) 2025-2026 Camilo Leal (camilolealdev) & Easy Marketing (https://easy-marketing.xyz)
Author / Maintainer: Camilo Leal (camilolealdev) — https://github.com/camilolealdev | https://easy-marketing.xyz
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction...
🇬🇧 This project is released under the MIT License — see LICENSE for the full text. The bundled third-party skills remain under their original licenses (referenced in each library's own files inside skills/).
🇪🇸 Este proyecto se publica bajo la Licencia MIT — ver LICENSE para el texto completo. Las skills de terceros agrupadas conservan sus licencias originales (referenciadas en los archivos de cada librería dentro de skills/).
🤝 Contributing · Seguridad
🇬🇧 See CONTRIBUTING.md (add skills, report issues, code style) · SECURITY.md (report vulnerabilities) · CODE_OF_CONDUCT.md.
🇪🇸 Ver CONTRIBUTING.md (añadir skills, reportar issues, estilo) · SECURITY.md (reportar vulnerabilidades) · CODE_OF_CONDUCT.md.
