PTENES
Claude Code skill

You build the foundation of your agentic OS

A personal coach that asks 2 or 3 questions in plain English and builds the actual files and folders for your own agentic operating system, layer by layer. You don’t need to know how to code — the skill handles the technical work; you just make the decisions.

Start the guide View on GitHub
OS Coach, the AI Systems Architect: you build the foundation. The six layers—Identity, Context, Rules, Skills, Tools, and Agents—make up the Agentic OS.
What it is

An agentic OS is mostly plumbing

The AI model is the cheap 20%. You can’t talk to your data if the plumbing is a mess — that’s why the skill sorts out the plumbing first, then builds capabilities on top.

🗣️ Speak naturally

Never sounds like a manual. No jargon — when a technical term is unavoidable, it gets a short sentence and an everyday analogy. Assumes you’ve never opened a terminal.

🧱 Build for real

It’s not a conversation: with each answer you give, it creates the actual files and folders in your folder, explaining what it did and why.

🔒 Protect sensitive information

If you mark a field as private, it never writes that value to the folder — it stores a reference that doesn’t identify you.

How it works

Six layers, built from the bottom up

Identity is the foundation, Substrate (the behind-the-scenes layer) is the largest layer, and Agents sit at the top — only after everything below them is solid.

1 · Identity→ 2 · Substrate→ 3 · Rules & Hooks→ 4 · Skills→ 5 · Tools→ 6 · Agents

❓ Ask

Asks 2 or 3 simple questions about the current layer, then stops and waits for your answer.

🏗️ Build

Creates the actual file or folder for that layer when you answer, tailored to your goal.

💾 Persist

Saves the decision and your place in memory.md, so nothing gets lost between sessions — and you can audit everything anytime with OS-AUDIT.md.

Prerequisites

That’s all you need

No dependencies, no API keys, no build step.

Claude Code

Installed on your machine.

# verify the installation
claude --version

The copied skill

Put the os-coach folder inside your Claude Code skills folder.

# install
cp -r os-coach ~/.claude/skills/

Any folder

Run it inside any folder you want to turn into an OS — that folder becomes the OS itself.

# inside the folder
mkdir meu-os && cd meu-os
User guide · step by step

From zero to your first agentic OS

You always take one small step at a time. The coach asks, builds, persists, and points you to the next layer.

1

Start with your goal

Clearly state in Portuguese where you want to get to. The skill confirms the goal and prepares the folder.

/os-coach start Eu vivo perdendo prazos de cliente e quero nunca mais atrasar uma entrega  # starts the OS in the current folder
2

Answer the questions for each layer

2 or 3 simple questions at a time — who this is for, what it should always do, what it should never do. It writes the Identity file (CLAUDE.md) in your own words.

# it writes something like
## Recusas inegociáveis (nunca)
- Eu nunca mando mensagem para um cliente sem você ler e aprovar antes.
3

Work through one layer at a time

Go to the next unfinished layer. It trains Substrate, Rules, Skills, Tools, and Agents, always reading from and writing to the same memory.md.

/os-coach next  # moves to the next layer and trains it
4

See where you are

At any time, check the progress map from memory, or jump straight to a specific layer.

/os-coach status                 # progress map
/os-coach layer skills           # skips to the Skills layer
5

Audit whenever you want

It scores all six layers against your goal, gives you the three highest-leverage next moves, and writes the report to OS-AUDIT.md.

/os-coach audit  # bulletin + next steps, in priority order
Examples

The six layers and the training loop

The complete Agentic OS map and the ask → build → persist → next cycle that feeds the audit.

The Agentic OS: six layers built from the bottom up: Identity, then Substrate, then Rules, then Skills, then Tools, then Agents
Six layers, from the bottom up: Identity is the foundation, Agents are at the top.
The training loop: ask, build, persist, next, all reading from and writing to memory.md, with the audit feeding back into the cycle
The training loop: ask → build → persist → next, with the audit feeding back into the cycle.
Roadmap

You build the foundation. It runs the agent.

OS Coach doesn’t handle your day-to-day tasks for you — it builds the foundation a reliable agent can run on.

Today
OS Coach builds the foundationIdentity, Substrate, Rules, Skills, Tools, and Agents — architecture, logic, and structure, without executing anything on its own.
Then
A runtime platform runs the agentOnce the foundation is solid, a platform like OpenClaw takes what was prepared here and puts it to work: messages, automations, tools, and dashboards.
Always
Ongoing auditsAsk for an assessment at any time, and OS Coach scores the six layers against your goal, feeding that back into your next steps.