Hermes /learn Command: Teach It Once, Keep It Forever

Share this post

The Hermes /learn command turns any source — a URL, a PDF, a folder, a job it just did — into a skill the agent writes itself and keeps forever. I fed it a guide from my own site and it worked first time. Here’s exactly how to use it.

Short answer

  • Type /learn plus a source; Hermes distils it into a reusable skill saved as skill.md.
  • No extra engine or installs — just update Hermes so the command exists.
  • Load by name with /skill, or let it auto-load on related trigger terms.
  • Works everywhere: dashboard, terminal, Jarvis voice and messaging.


What the Hermes /learn command does

You point Hermes at a source — a URL, a document, a PDF, a folder of code, pasted notes, even a job it just finished — and type one word: /learn. Hermes reads it, distils it into a verifiable, reusable skill, and saves it for every future session as a skill.md markdown file.

The demo that sold me: I fed it a guide from my own website about the agent operating system — just /learn plus the URL in chat — and it came back with “learned it and saved it for future sessions”: what the system is, the six pillars, the main surfaces, the setup path. First time, no retries.

There’s no separate learning engine and no extra install. It uses the tools Hermes already has — just make sure you’re on the latest version (hermes update) so the command exists.

Why the /learn command matters

Every session used to start with re-explaining. You’d walk the agent through a workflow, it would do the job, then forget the whole recipe the moment you were done — you were the only one who remembered how your workflows run.

Now you show it once. It writes itself a clean, reusable recipe, keeps it, and next time you just say do the thing. Five minutes once instead of a few minutes every single time — and it compounds: every skill you teach makes the agent more capable for good.

🔥 Want this set up without the guesswork? The full /learn tutorial — turning your daily workflows into kept skills — is on the roadmap step by step. Inside the AI Profit Boardroom you get the Agent OS as a ready-to-install file, a 30-day roadmap, daily tutorials the same day new tools ship, and four live coaching calls a week where you share your screen and get unstuck. → Get access here

Using Hermes /learn, step by step

  1. Update Hermes so the command is available.
  2. Open a chat and point at a source/learn plus a URL, a file, a folder, or a plain-English description of the workflow you just ran.
  3. Let it write the skill. Hermes reads the source and authors a clean skill.md on its own.
  4. Approve the write — if your approval gate is on, it asks before saving; you see the file before it’s kept.
  5. Use it forever. Ask for the task by name, load it explicitly with /skill <name>, or just mention a related term and it loads itself.

Two quality levers from real use: a clearer source gives a better skill, and for tricky jobs you can route /learn through a stronger model. Though for scale — I watched a free local model write a clean skill in about 25 seconds.

The /learn command works everywhere

One learned skill follows the agent across every surface: the dashboard chat, the terminal, voice through Jarvis, and the messaging apps. Teach it once at your desk; use it by voice on a walk.

It’s also different from bolt-ons in one important way. Downloaded skills and MCP servers are other people’s tools; /learn captures how you do things — your steps, your rules, the stuff you can’t download anywhere. What the file actually contains, when it auto-loads, and how to fix a bad one is covered in the /learn skill breakdown — and the wider skill strategy in Hermes SEO skills.

FAQ

What is the Hermes /learn command?

One command that turns any source — URL, PDF, folder, notes, or a job just finished — into a reusable skill Hermes writes itself and keeps forever as a skill.md file.

How do I use /learn?

Update Hermes, then type /learn plus your source in any chat. It authors the skill, you approve the write, and it’s saved for every future session.

Do I need to write any code?

No. Plain English is fine — Hermes writes the skill file itself. You never touch a line unless you want to edit one.

Does it need a special setup?

No separate engine, no extra installs — it uses the tools Hermes already has. Just be on the latest version.

Where does /learn work?

Everywhere the agent lives: dashboard, command line, Jarvis voice, and messaging — one skill, every surface.

How is this different from installing someone’s skill?

Installed skills are generic tools. /learn captures your own workflow as a recipe — the part you previously had to re-explain every session.

The bottom line on the Hermes /learn command

The Hermes /learn command is teach-once-keep-forever in one word: point it at anything, let it author its own skill.md, approve the write, and stop re-explaining yourself to a tool. Every week you use it, the gap grows between your agent and a fresh install — start with the workflow you repeat most.

About Julian Goldie

I run Goldie Agency, a 7-figure SEO agency, and teach this daily on a 400K+ subscriber YouTube channel. 240+ client projects on Upwork at a 100% job-success score, 10+ years through every major Google update. My systems are in the AI Profit Boardroom; my link building book is free here.

Related reading

Last updated August 2026. This is the living guide to Hermes /learn command — it gets updated as the tools change.

Table of contents

Related Articles

Hermes /learn: Hermes /learn is the learn-anything engine — show the agent something once and it keeps the skill forever. Here’s what it
Claude Code desktop DeepSeek: Wiring a Claude Code desktop DeepSeek setup just went official: Ollama now ships Claude Desktop support as a third-party
Hermes /learn skill: Hermes learns from any guide you give it and keeps it forever as a skill.md — what gets saved, when it loads, how to fix or share it.
DeepSeek Harness agent: A DeepSeek Harness agent is what you get when you give DeepSeek’s body a brain: it reads files, writes across them, runs