A curated, production-tested gallery of open-source agent skills, system protocols, and operational blueprints. Built by Grupo Plus Tech for developers, sysadmins, and amateur radio operators who orchestrate AI agents across different execution environments.
Every skill in this repository is designed as multi-runtime, providing ready-to-use definitions for:
- 🦙 Hermes Agent (
SKILL.md) - 🤖 Claude / Claude Code / Projects (
CLAUDE.md) - ⚡ Cursor IDE (
CURSOR.mdc) - 💎 Google Gemini AI Studio (
GEMINI.md) - 💬 ChatGPT / Custom GPTs (
GPT.md)
skills/
├── 📻 ham-radio/
│ ├── dmr-linux-codeplug/ # Native Linux DMR programming (qdMR, QLog, Codeplug vs User DB)
│ ├── hf-digital-modes/ # HF Digital Modes (FT8, JS8Call, CAT Control CI-V / Rigctl)
│ └── space-weather-hf-propagation/ # Real-time NOAA solar metrics (SFI, Kp, flares) & HF band forecast
├── 🛠️ agent-ops/
│ ├── agent-state-backup/ # Safe Git versioning & DR snapshots for live AI agents without leaks
│ ├── hermes-subagent-pattern/ # Multi-agent orchestration pattern (profiles, delegation, lifecycle)
│ ├── stdio-mcp-server-builder/ # Zero-dependency stdio MCP server builder in Node.js & Python
│ └── vps-health-snapshot/ # Single-run server diagnostics (Docker, SSL, Ports, Cron, RAM)
├── 🌐 web/
│ ├── web-accessibility-audit/ # WCAG 2.2 AA accessibility auditor for Next.js/React & contrast math
│ └── nextjs-secure-actions/ # Security hardening for Next.js Server Actions (IDOR, Zod, auth)
└── 🔌 integrations/
├── telegram-supergroup-topics/ # Telegram Forum Topics auto-discovery and scheduled thread delivery
├── composio-sdk-quirks/ # Diagnostics for Composio SDK 1.0+ project mismatch & MCP routing
└── nextjs-protected-screenshots/ # Headless session-injected Playwright audits for protected Next.js apps
| Category | Skill | Functions / Capabilities | Runtime Compatibility |
|---|---|---|---|
| Ham Radio | dmr-linux-codeplug |
Codeplug architecture, qdMR setup, OpenGD77/Baofeng/AnyTone quirks, QLog sync, udev rules | Hermes, Claude, Cursor, Gemini, GPT |
| Ham Radio | hf-digital-modes |
WSJT-X, JTDX, CI-V protocol debugging, ALSA audio loopbacks, ALC & PTT troubleshooting | Hermes, Claude, Cursor, Gemini, GPT |
| Ham Radio | space-weather-hf-propagation |
NOAA SWPC live API ingestion, Planetary Kp, Solar Flux (SFI), ionospheric MUF band forecast | Hermes, Claude, Cursor, Gemini, GPT |
| Agent Ops | agent-state-backup |
Safe .gitignore patterns, Git automated sync, zero secret leaks, SQLite WAL lock prevention |
Hermes, Claude, Cursor, Gemini, GPT |
| Agent Ops | hermes-subagent-pattern |
Isolated personas, state boundaries, delegation loops, background monitor patterns | Hermes, Claude, Cursor, Gemini, GPT |
| Agent Ops | stdio-mcp-server-builder |
JSON-RPC 2.0 framing, stdin/stdout buffering, stream isolation (logs to stderr), tool schemas | Hermes, Claude, Cursor, Gemini, GPT |
| Agent Ops | vps-health-snapshot |
Unified terminal audit: Docker status, open listening ports, SSL cert expiry, cron jobs | Hermes, Claude, Cursor, Gemini, GPT |
| Web Dev | web-accessibility-audit |
WCAG 2.2 AA contrast formulas, keyboard navigation, semantic landmarks, ARIA audit | Hermes, Claude, Cursor, Gemini, GPT |
| Web Dev | nextjs-secure-actions |
Audits 'use server' actions against authorization bypass, IDOR vulnerabilities & Zod validation |
Hermes, Claude, Cursor, Gemini, GPT |
| Integrations | telegram-supergroup-topics |
Extract hidden Telegram thread_id, route scheduled messages and crons to specific topics |
Hermes, Claude, Cursor, Gemini, GPT |
| Integrations | composio-sdk-quirks |
Fix zero-account listing bug, project mismatch, 401 vs 502 triage in Composio Tool Router | Hermes, Claude, Cursor, Gemini, GPT |
| Integrations | nextjs-protected-screenshots |
Automated visual QA with Playwright injecting auth cookies and bypass tokens in Next.js apps | Hermes, Claude, Cursor, Gemini, GPT |
(Inspirado en el análisis de Fazt Code y Dax de OpenCode sobre la saturación de contexto)
Muchos repositorios populares de skills cometen el error de empujar mega-packs de 50+ archivos Markdown. Esto degrada el rendimiento de los modelos modernos por tres razones críticas:
- Contaminación de Contexto (Trigger Polución): Cada skill registrado gasta tokens en el arranque solo con su descripción. Si la descripción es ambigua, el agente lo auto-invoca sin necesidad, llenando la ventana de contexto de texto estático.
- Restricciones que Limitan la Inteligencia del Modelo: Forzar reglas rígidas pensadas para modelos antiguos ahoga la capacidad de modelos avanzados (Claude 3.7, Gemini 2.5, GPT-4.5) que ya resuelven el problema de forma nativa ("vainilla").
- Falsas Métricas de Estrellas: Decenas de miles de estrellas en GitHub no garantizan utilidad en tu terminal diario.
- Instalación Granular, No Bundles Ciegos: Instala únicamente el skill que resuelve tu problema actual en tu proyecto.
- Ingeniería y Protocolos, No Prose Prompting: Nuestros skills no dicen "sé un programador experto"; entregan comandos reproducibles, endpoints de NOAA, fórmulas de WCAG, udev rules y protocolos de hardware.
- Descripciones con Triggers Precisos: Cada skill define exactamente cuándo activarse y cuándo mantenerse en reposo.
See INSTALL.md for full instructions tailored to each platform.
# Clone directly into your Hermes skills directory
cd ~/.hermes/skills
git clone https://github.com/grupoplustech/skills.git grupoplustech-skillsHermes will automatically index every SKILL.md in the directory.
Symlink or copy the CLAUDE.md from the desired skill into your project root or ~/.claude/skills/.
Copy CURSOR.mdc into .cursor/rules/<skill-name>.mdc in your workspace.
Copy the contents of GEMINI.md or GPT.md directly into the System Instructions / Instructions field of your agent.
We welcome additions, fixes, and community contributions!
- Want to contribute a new skill? Check CONTRIBUTING.md.
- Have a radio, web, or agent infrastructure pattern that solved a tricky bug? Send a Pull Request!
- Please ensure you include all multi-runtime files (
SKILL.md,CLAUDE.md,CURSOR.mdc,GEMINI.md,GPT.md) and redact any private API keys or personal identifiers.
Distributed under the MIT License. Free for commercial and personal use. Maintained with ❤️ by Grupo Plus Tech.