For Hermes Agent, provide an out-of-the-box set of skills and workflow layer covering the full application lifecycle from idea to deployment to operations. https://github.com/Salomondiei08/oh-my-hermes… Oh My Hermes is like installing an “operating system” for Hermes — once installed, Hermes is no longer just a chatbot but a capable ops + development partner. 20 skills cover requirement clarification, product brief, design handoff, engine selection, code implementation, deployment, database connection, monitoring, notifications, GitHub operations, and more. 5 agents (CTO, PM, Dev, QA, Ops) each manage their own segment, collaborating on a kanban board. Hermes itself is the primary operator, while Claude Code and Codex are optional deep coding accelerators.

Salomondiei08/oh-my-hermes

Source: https://github.com/Salomondiei08/oh-my-hermes

Oh My Hermes

Stars (https://github.com/salomondiei08/oh-my-hermes/stargazers)
License
Hermes (https://hermes-agent.nousresearch.com)

Skills Agents

An opinionated workflow layer for building, shipping, and operating apps — delivered directly to Hermes.

Like Oh My Zsh is to Zsh. You install it once and Hermes becomes genuinely useful for real software projects. Not a chatbot wrapper. Not a prompt pack. A curated set of skills that Hermes loads and runs autonomously — on your VPS, on your laptop, wherever Hermes lives.

AI agent? See INSTALL_FOR_AGENTS.md for the one-command install protocol.

Use with an AI agent

If you work with Claude, Cursor, Copilot, or any coding assistant, paste this directly into the chat:

To install Oh My Hermes on a project:

`` Install Oh My Hermes on this project.

1. Clone the repo: git clone https://github.com/salomondiei08/oh-my-hermes /tmp/oh-my-hermes
2. Run the installer: bash /tmp/oh-my-hermes/install.sh
3. Bootstrap this project: bash /tmp/oh-my-hermes/scripts/bootstrap.sh
4. Verify everything installed: bash /tmp/oh-my-hermes/scripts/verify.sh Then tell me what’s missing and what I need to fill in. ``

To set up the full autonomous CTO loop:

`` Set up the Oh My Hermes CTO loop for this project. Follow the instructions in: https://raw.githubusercontent.com/salomondiei08/oh-my-hermes/main/INSTALL_FOR_AGENTS.md I will provide the following when you ask:

• My GitHub repo (owner/repo)
• A GitHub fine-grained token (I’ll create one if you explain how)
• My production URL (if deployed) Walk me through each step one at a time. ``

To start a new app from scratch:

`` Use Oh My Hermes to start a new app. Oh My Hermes is installed at ~/.hermes/skills/. Load the following skills in order:

1. clarify-requirements — ask me the 7 questions and save my answers
2. product-brief — generate the brief from my answers
3. choose-engine — decide how to implement it
4. implement — build it using the right engine Start with clarify-requirements now. ``

The core idea

Hermes is the operator. It talks to you. It remembers. It builds. It deploys. It monitors. It notifies. You describe what you want in plain language — on Telegram, Slack, Discord, your terminal, wherever you have Hermes configured. Hermes loads the right skill and runs the workflow.

Claude Code and Codex are optional engines Hermes can invoke when a task needs deep file editing — but Hermes handles the orchestration, the ops, the memory, and the lifecycle on its own.

Hermes does not need Claude Code or Codex to be useful. Those are optional. Hermes itself has a terminal backend and can write, edit, and run code directly.

How it all fits together

┌─────────────────────────────────────────────────────────────────┐ │ YOU (founder) │ │ Telegram · Slack · Discord · WhatsApp · terminal │ └──────────────────────────┬──────────────────────────────────────┘ │ plain-language messages ▼ ┌─────────────────────────────────────────────────────────────────┐ │ HERMES (VPS / local, 24/7) │ │ │ │ ┌──────────┐ routes to ┌─────────────────────────────────┐ │ │ │ Gateway │ ────────────▶ │ CTO Agent │ │ │ │(Telegram │ │ monitors kanban, orchestrates │ │ │ │ /Slack...│ └───────────┬─────────────────────┘ │ │ └──────────┘ │ spawns sub-agents │ │ │ │ │ ┌──────────┼──────────────┐ │ │ ▼ ▼ ▼ │ │ ┌────────────┐ ┌──────────────┐ ┌──────────┐ │ │ │ PM · Dev │ │ QA · Security│ │ Ops │ │ │ │ triage │ │ review · scan│ │ deploy │ │ │ │ implement │ │ approve │ │ monitor │ │ │ └─────┬──────┘ └──────┬───────┘ └────┬─────┘ │ │ │ │ │ │ │ ┌──────▼────────────────▼───────────────▼──────┐ │ │ │ Hermes Kanban │ │ │ │ Backlog → In Progress → Review → Done │ │ │ └───────────────────────────────────────────────┘ │ │ │ │ Persistent memory · 23 skills · 5 workflows · cron jobs │ └──────┬──────────────────────────────────────────────────────────┘ │ ├──▶ Claude Code (complex multi-file coding) ├──▶ Codex (quick single-file fixes) ├──▶ Vercel (hosting + preview URLs) ├──▶ Supabase (database + auth + migrations) ├──▶ GitHub (issues, PRs, merge) └──▶ Sentry / Uptime Kuma (monitoring)

Full project lifecycle

From idea to running production app, Hermes handles each stage:

IDEA ↓ onboarding ← Bot asks everything in chat, configures the loop ↓ clarify-requirements ← Hermes asks 7 questions, saves answers to memory ↓ product-brief ← Writes PRODUCT_BRIEF.md from requirements ↓ design-handoff ← Converts your design notes to an implementation spec ↓ choose-engine ← Decides: Hermes / Claude Code / Codex ↓ implement ← Builds it — surgical changes, no secrets committed ↓ deploy-to-vercel ← Pre-deploy checks → deploy → captures URL ↓ connect-supabase ← Links DB, pushes migrations, sets env vars ↓ setup-monitoring ← Sentry + Uptime Kuma ↓ post-deploy-followup ← Health check → log → notify you ↓ RUNNING APP ↓ auto-issue-triage ← Every hour: scores issues, picks top priority ↓ implement + PR ← Dev Agent builds, Security Agent reviews ↓ await-merge-approval ← Sends you YES/NO message ↓ YOU reply YES ← Merges, deploys, health-checks, confirms ↓ REPEAT — Hermes watches it, you approve changes

The autonomous CTO loop

Once configured, this runs every hour without you touching anything:

GitHub issue opens ↓ PM Agent scores & triages → kanban: Backlog ↓ Dev Agent implements → kanban: In Progress ↓ Security Agent: secret scan + OWASP check + CVE check ↓ QA Agent: build check + health check + plain-English summary ↓ YOU get a message on Telegram: ──────────────────────────────────────── PR #12 — Fix login redirect What changed: Users who hadn't verified their email were sent to a broken page. They now see "Check your inbox." Build: passing | Preview: healthy (180ms) | No secrets found Preview: https://myapp-fix-login.vercel.app Reply YES to ship. Reply NO and tell me why. ──────────────────────────────────────── ↓ YES → merges, deploys, health-checks, confirms live URL to you NO → Dev Agent iterates on your feedback

Get started

Step 1 — Install Hermes Agent
Follow the Hermes quickstart (https://hermes-agent.nousresearch.com/docs/getting-started/quickstart). At the end you have a bot you can message on Telegram (or Slack, Discord, WhatsApp).

Step 2 — Install Oh My Hermes
bash curl -fsSL https://raw.githubusercontent.com/salomondiei08/oh-my-hermes/main/install.sh | bash

Step 3 — Message your bot
set up the CTO loop The bot will ask for your GitHub repo, walk you through creating a token step by step, ask for your production URL, and configure everything. No terminal needed after Step 2.

What you unlock at each step:

Skills

Agents

Role definitions live in agents/. Running scripts/setup-cto.sh (or messaging “set up the CTO loop”) creates the actual Hermes profiles and makes the agents active.

Workflow examples

Start a new project: you: start a new app hermes: What problem does this solve? Who experiences it? you: [answer] hermes: [6 more questions...] hermes: Requirements saved. Generating product brief... hermes: Brief written to PRODUCT_BRIEF.md. Ready to implement or do design first?

Deploy after implementing: you: deploy this to Vercel hermes: Running pre-deploy checklist... hermes: Deploying... done. URL: https://myapp.vercel.app hermes: Health check: PASS (200ms) hermes: Notification sent to Slack.

Quick fix: you: fix the auth redirect bug in src/middleware.ts hermes: Loading context... routing to Codex (single-file fix) hermes: Done. Typecheck passes. Creating PR... hermes: PR #14 ready — reply YES to ship.

Steer mid-session (Hermes v0.13+): you: /steer prioritize the payment bug above everything else hermes: Understood. Switching Dev Agent to issue #38.

Default stack

All pluggable. Each skill documents how to substitute.

Running on a VPS

The intended setup for production use — Hermes runs 24/7, crons fire automatically, you interact from your phone:

``bash

On a $5/month VPS (Ubuntu 22.04+)

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash hermes model # choose your provider (Anthropic, OpenAI, etc.) hermes gateway setup && hermes gateway start # connect Telegram or Slack

Then install Oh My Hermes

curl -fsSL https://raw.githubusercontent.com/salomondiei08/oh-my-hermes/main/install.sh | bash

Message your bot: “set up the CTO loop”

``

For Docker: bash docker run -d --restart=always \ -v hermes-data:/root/.hermes \ nousresearch/hermes-agent

Installation scripts

Optional: GBrain memory backbone

GBrain (https://github.com/garrytan/gbrain) gives Hermes a richer, self-updating knowledge graph — people, companies, decisions, deployment history — queryable across sessions.

bash git clone https://github.com/garrytan/gbrain.git ~/gbrain && cd ~/gbrain curl -fsSL https://bun.sh/install | bash && export PATH="$HOME/.bun/bin:$PATH" bun install && bun link && gbrain init

Do not use npm install -g gbrain — a squatter package exists on npm under that name.

Architecture

oh-my-hermes/ ├── skills/ ← 23 skill files → ~/.hermes/skills/ ├── workflows/ ← 5 workflow files → ~/.hermes/workflows/ ├── agents/ ← 6 agent role definitions → ~/.hermes/agents/ ├── templates/ ← AGENTS.md template, .env example, health endpoint ├── scripts/ ← install, bootstrap, verify, setup-cto, uninstall └── docs/ ← Full documentation

See docs/architecture.md for detail.

Roadmap

V1 — current
23 skills, 6 agents, 5 workflows, chat-guided onboarding, security agent, Karpathy code principles, one-command CTO setup, Vercel + Supabase + GitHub ops.

V2 — planned
Rollback skill, staging → production promotion workflow, incident creation, post-deploy automated tests.

V3 — planned
Multi-service orchestration, more example apps, hosted setup wizard.

Star history

Star History Chart (https://star-history.com/#salomondiei08/oh-my-hermes&Date)

Contributing

Read docs/architecture.md before proposing features.
Open issues for wrong or missing skills, bugs in scripts, or Hermes improvement proposals.

License

MIT