The same system that runs a real web design agency, now with 545 production skills extracted from a live 7-agent fleet.
This is the exact configuration, tools, and skills extracted from running Weblyfe: a web design agency powered by AI employees (Appies) that handle:
- 📧 Email triage and responses
- 📅 Calendar management
- 🎯 Lead capture and CRM automation
- 📝 Proposal generation and tracking
- 🌐 Website builds and deployments
- 📊 Competitor monitoring
- 🔒 Security scanning
- 💬 Customer support across Telegram, WhatsApp, Discord
This is not a chatbot. This is a digital team member that works 24/7, learns from mistakes, and gets stronger every day.
Released June 2026.
- 545 production skills extracted from the live Appie fleet (Mac Mini, GPU/media workers, VPS nodes, and approved client-bot hosts). Up from 156 in v4.4.
- Hermes Agent compatibility across the board. Skills were originally OpenClaw-only; the fleet now runs primarily on Hermes Agent.
- MiniMax M2.7 as the default model via OpenRouter: 17x cheaper than Claude at GPT-4 class quality.
- fal.ai + RunPod media stack for image and video generation (Kling, Nano Banana, ComfyUI workers).
- UI/UX Pro Max design framework: 67+ UI styles, 161 palettes, 57 font pairings, 99 UX guidelines bundled as a skill.
- 9-category skill layout: automation, communication, content, integrations, knowledge, meta, ops, personal, and ecc.
Full setup guide: Build Your Own Techwiz PDF v4.5
- Hermes Agent installed (or OpenClaw)
- An OpenRouter API key (MiniMax M2.7 recommended, ~$0.01/1M tokens)
- A Telegram bot token (optional, for messaging interface)
The default for v4.4 is MiniMax M2.7 via OpenRouter:
- Cost: ~$0.01/1M tokens (vs $3.50/1M for Claude Opus)
- Quality: equivalent to GPT-4 class
- Setup: get an OpenRouter key at openrouter.ai
git clone https://github.com/S3YED/appie-kit.git
cd appie-kit
./install.sh /path/to/your/workspace/nano workspace/SOUL.md # Who your AI is: personality, values, communication style
nano workspace/USER.md # Who you are: name, timezone, preferences, context
nano workspace/TOOLS.md # What tools it can access: API references, credentialscp .env.example .env.secrets
chmod 600 .env.secrets
nano .env.secrets # Fill in your API keysFor Hermes Agent:
cp -r skills/content/design ~/.hermes/skills/
cp -r skills/knowledge/exa-plus ~/.hermes/skills/
# or copy entire categories:
cp -r skills/content/* ~/.hermes/skills/For OpenClaw:
cp -r skills/content/design ~/.openclaw/skills/# Hermes Agent
hermes start
# OpenClaw
openclaw gateway startThat's it. Your AI employee is live.
545 public skills across 9 categories. Each category has its own index.
| Category | Net Skills | What it covers |
|---|---|---|
| automation | 27 | Agent self-management, fleet coordination, brain sync, sub-agent spawning, ClawFlow workflows |
| communication | 6 | Telegram, Discord, Slack, iMessage (BlueBubbles), WhatsApp (wacli), voice calls |
| content | 59 | UI/UX design, image gen, video production, SEO, copywriting, Next.js, TTS, stable diffusion |
| integrations | 109 | Airtable, Notion, Google Workspace, GitHub, n8n, fal.ai, Spotify, Webflow, Trello, Gemini |
| knowledge | 33 | LLM fine-tuning (LoRA/GRPO), quantization, vLLM, Whisper STT, DSPy, model evaluation |
| meta | 11 | Skill authoring, ClawHub registry, task planning, dogfood QA, brainstorming frameworks |
| ops | 28 | GitHub/git, DigitalOcean, tmux, 1Password, healthcheck, session management, fleet maintenance |
| personal | 22 | Bear Notes, Spotify, Things 3 (macOS), weather |
| ecc | 250 | Imported MIT engineering and agentic-ops skills |
545 ready-to-drop-in skills across 9 categories. The full skill library from the live fleet is larger; this kit contains the skills that transfer cleanly to any Hermes or OpenClaw setup.
| Skill | Description |
|---|---|
| frontend-design | Production-grade UI (React, HTML/CSS, Next.js, Vue). Bold typography, motion systems, gradient meshes |
| ui-ux-pro-max | Design decision engine: 67+ UI styles, 161 palettes, 57 font pairings, 99 UX guidelines |
| tips-landing-pages | Conversion-optimized landing pages using the TIPS framework |
| 21st-dev | Install shadcn/ui-compatible React components from 21st.dev registry |
| design | Logo generation (55 styles), CIP mockups (50 deliverables), banners, icons, social |
| design-mastery | Core design knowledge: hierarchy, spacing, typography, color, depth, UX laws |
| design-system | Token architecture, three-layer tokens (primitive/semantic/component), CSS variables |
| ui-styling | shadcn/ui + Tailwind CSS. Accessible components, dark mode, responsive layouts |
| banner-design | Multi-format banners across 22 art styles |
| brand | Brand identity, voice, messaging frameworks, design tokens |
| slides | Strategic HTML presentations with Chart.js and design tokens |
| Skill | Description |
|---|---|
| nextjs-expert | Next.js 15 App Router specialist. Server Components, auth, caching, streaming |
| coding | Code quality guidelines, patterns, security rules, debugging workflow |
| gitclaw | Automated workspace backup to GitHub via crontab |
| architecture-diagram | Generate system architecture diagrams from descriptions |
| github | GitHub repository management, PR review, issue tracking |
| Skill | Description |
|---|---|
| agentic-video-tools | Compare and automate agentic video editing APIs (Vizard, Submagic, Descript) |
| video-editing-pro | Cut-decision framework for short-form video. Hook precision, energy curve, captions |
| seo-checklist | Pre-publish on-page SEO checklist: title, meta, H1, images, breadcrumbs |
| content-creation | Structured content creation across formats and platforms |
| Skill | Description |
|---|---|
| digital-ocean | Manage DigitalOcean droplets, domains, infrastructure via DO API |
| n8n | Interact with n8n workflows: list, trigger, monitor, manage automations |
| healthcheck | Fleet health monitoring and alerting |
| web-scraping-javascript-sites | Scrape SPA/React sites. urllib-first strategy, Playwright fallback, stealth mode |
| Skill | Description |
|---|---|
| exa-plus | Neural web search via Exa AI. People, companies, news, research, code |
| read-github | Read GitHub repos via gitmcp.io. Semantic search, zero hallucination on structure |
| Skill | Description |
|---|---|
| clawflow | Multi-step task execution spanning detached background jobs |
| kanban-orchestrator | Decomposition playbook and specialist-roster conventions for parallel work |
| dispatch-multiple-agents | Spawn 2+ independent agents in parallel for independent tasks |
| brain-sync | Sync knowledge and learnings to the shared agent-brain repo |
| memory-search | Search and retrieve from the agent brain: daily logs, project notes, topics |
appie-kit/
├── workspace/ # Core agent identity and behavior files
│ ├── SOUL.md # Personality, values, communication style
│ ├── USER.md # Who you are: timezone, preferences, context
│ ├── TOOLS.md # Available tools and API references
│ ├── AGENTS.md # Operating rules, safety, group chat behavior
│ ├── IDENTITY.md # Multi-agent identity (for fleets)
│ ├── HEARTBEAT.md # Proactive check-in configuration
│ └── memory/ # Persistent memory directory
├── skills/ # 545 public skills, 9 categories
│ ├── automation/ # Agent self-management, fleet coordination
│ ├── communication/ # Messaging platform integrations
│ ├── content/ # Design, video, SEO, copywriting, frontend
│ ├── integrations/ # Third-party APIs and SaaS connectors
│ ├── knowledge/ # ML/AI fine-tuning, search, research
│ ├── meta/ # Skill authoring, planning, QA
│ ├── ops/ # Infrastructure, git, DevOps
│ ├── personal/ # Notes, tasks, music (macOS)
│ └── INDEX.md # Master skill index with counts
├── tools/ # Production shell scripts
│ ├── validate-public-skills.py # Pre-publish skill validation
│ ├── setup-openclaw-mac.sh # Full Mac setup
│ ├── setup-openclaw-vps.sh # VPS setup (Ubuntu/Debian)
│ ├── hermes-agent-install.sh # Install Hermes Agent on VPS
│ ├── security-scan.sh # Scan for exposed secrets
│ ├── health-check.sh # Fleet health monitoring
│ ├── session-manager.sh # Clean stale sessions
│ └── safe-gateway-restart.sh # Zero-downtime gateway restart
├── configs/ # Platform and fleet configuration examples
│ ├── fleet-access.example.yml # Private fleet-access template
│ ├── telegram.example.yml
│ ├── discord.example.yml
│ ├── whatsapp.example.yml
│ └── multi-agent.example.yml
├── prompts/ # Production-tested prompt library
├── case-studies/ # Real-world examples with numbers
├── install.sh # One-command installer
├── .env.example # Environment variables template
├── CONTRIBUTING.md # How to add skills and contribute
├── CHANGELOG.md # Version history
└── LICENSE # MIT
| Category | Tool | Why |
|---|---|---|
| AI model | MiniMax M2.7 (OpenRouter) | 17x cheaper than Claude, GPT-4 class quality |
| Agent framework | Hermes Agent (Nous Research) | Open source, persistent memory, multi-platform |
| Alternative framework | OpenClaw | macOS-native, good for desktop automation |
| Messaging | Telegram | Best bot API, instant delivery, group support |
| CRM | Airtable | Flexible, API-first, automatable |
| Automation | n8n | Self-hosted, 400+ integrations |
| Hosting | Vercel | Zero-config deploys, edge functions |
| VPS | DigitalOcean | Simple API, affordable |
| Networking | Tailscale | Zero-config VPN for fleet communication |
| Image gen | fal.ai (Nano Banana) | Fast, cost-effective, consistent results |
| Video gen | fal.ai (Kling 3.0) | Best character consistency, 4K |
| Search | Exa | AI-native search API, better than Google for research |
| Desktop agent | Clawd Cursor | Screen control, universal app automation |
Before: manual email checks, copy-paste to spreadsheet, manual follow-ups. After: Form submission triggers Airtable record, Brevo list entry, confirmation email, and timed follow-up sequence. Result: 0 manual steps, under 2 seconds end-to-end, 24/7.
Appie-1 (Orchestrator, Mac Mini), Appie-2 (Marketing, VPS), Appie-3 (DevOps, VPS). Cost: ~$148/month for 3 AI employees running continuously. Setup: configs/multi-agent.example.yml
Input: one brand brief or topic idea. Output: AI video (Kling 3.0) + voiceover + social posts + blog draft. Time: 15 minutes vs 4+ hours manual.
Problem: every SaaS tool needs its own API integration, keys, auth flows, rate limits. Solution: Clawd Cursor gives your Appie eyes and hands. It sees your screen and controls your cursor. Result: one skill replaces dozens of API integrations. If you can click it, your agent can too.
- Always Be Kind: Your AI should be warm and respectful
- Always Help: Go beyond the minimum, anticipate needs
- Value Life and Humanity: AI serves humans, not the other way around
- Spread Positivity: Optimism is a strategy
- Expand Abundance: Create more than you consume
This repo contains no secrets. All API keys use placeholder values.
Before going live:
- Review
workspace/SOUL.mdfor any personal information - Verify
.env.secretshas no real keys committed - Run
tools/security-scan.shafter setup - Run
python3 tools/validate-public-skills.pybefore publishing changes - Set file permissions:
chmod 600 .env.secrets - Use Tailscale for remote access, never expose SSH publicly
- Keep filled fleet access files private. Use
configs/fleet-access.example.ymlonly as a placeholder template
- Getting Started: 20-minute walk-through from clone to first agent response
- Fleet Access: safely add worker/client-bot hosts for skill inventory without leaking private data
- FAQ: setup, skills, models, privacy, hardware, use cases
- Troubleshooting: symptom-first index for common issues
- Skills Catalog: browse 545 skills by category
- Contributing: how to add skills and submit changes
- Changelog: version history
- Code of Conduct: community standards
- Security Policy: how to report a vulnerability
- v4.5 guide: Build Your Own Techwiz PDF
- Hermes Agent: github.com/NousResearch/hermes-agent
- OpenClaw: github.com/openclaw/openclaw
- MiniMax M2.7 via OpenRouter: openrouter.ai
MIT: Use it, modify it, build on it. Just don't blame us if your AI orders 1000 pizzas.
Built by Seyed Hosseini at Weblyfe.
Powered by Hermes Agent and OpenClaw.
"From doctor to automation architect. If I can do it, you can too."