120+
Skills
15
Domains
Extensible
📖 How to Use & Extend
🃏 All Skills

🧠 What is Hermes?

Hermes is a persistent AI agent that lives across sessions. Unlike a one-shot chatbot, it remembers you — preferences, project context, past conversations — and executes complex multi-step workflows using a growing library of skills (reusable procedures). It runs on your server 24/7, can schedule jobs, browse the web, write and run code, talk to external APIs, and spawn sub-agents to parallelize work.

💬 How to Talk to Hermes

  • Speak naturally — no special commands needed
  • Hermes searches its skill library automatically
  • Give goals, not instructions: "Deploy my site" not "Run git push"
  • Provide context upfront to save iterations
"Schedule a weekly blog post ideas digest every Monday at 9am and post it to Slack"
"QA test flyfishingxl.com on mobile and fix any layout bugs"

⚡ Trigger Skill Superpowers

  • Hermes auto-loads relevant skills per task
  • Skills contain expert knowledge + API quirks
  • You can ask to use a specific skill by name
  • Skills stack — agent + browser + review combined
"Use the gstack-qa skill to test my site"
"Run a CEO-mode plan review on this feature spec"

📅 Scheduling (Cron Jobs)

  • Create recurring jobs with natural language
  • Results delivered to Slack, Telegram, email, etc.
  • Chain jobs (output of A feeds into B)
  • Jobs are persistent — survive restarts
"Every morning at 7am, check ArXiv for new LLM papers and summarize the top 3 for me in Slack"

🔌 Adding New Skills

  • After a complex task succeeds, say "Save this as a skill"
  • Skills live in ~/.hermes/skills/
  • Format: YAML frontmatter + Markdown body
  • Include: trigger conditions, steps, pitfalls, verification
  • Hermes auto-patches skills when it finds gaps
"Save how you just deployed my site as a skill for next time"

🤖 Sub-Agents & Delegation

  • Hermes can spawn parallel sub-agents for complex tasks
  • Claude Code, Codex, OpenCode — full coding agents
  • GStack team: 46 specialized virtual engineer roles
  • Sub-agents work independently, results come back to you
"Have Claude Code refactor my entire API layer while you plan the next feature"

🔮 What's Possible

  • Full-stack web development end-to-end
  • Fine-tune & deploy local LLMs
  • Automate any repetitive workflow
  • Research + synthesize + publish
  • Monitor & alert on external changes
  • Control smart home, music, media
  • Build and evolve its own skill library

🗂 Skill Building Standard Power User

Every skill follows a standard format for maximum effectiveness: a budget block (token limits), a state block (phase tracking), a phase dispatch table (what to do when), and sub-agent briefs under 600 tokens. The full standard is in the skill-building-standard skill. When a skill you use is missing steps or has wrong commands, Hermes patches it automatically.