Agent SkillOS

The operating system for self-improving AI agents.

Every job teaches the network. Every approved skill upgrades every authorized Agent. One Agent learns, all Agents level up.

This repository is preconfigured for:

GitHub organization: MontrealAI
Repository name: skillos
Live website: https://montrealai.github.io/skillos/
Deployment: GitHub Actions → GitHub Pages

Nontechnical start

Open this first:

README_FIRST_GITHUB_WEB_USERS.md

The short version:

1. Create MontrealAI/skillos on GitHub.
2. Upload the CONTENTS of UPLOAD_THE_CONTENTS_OF_THIS_FOLDER_TO_GITHUB.
3. Go to Settings → Pages → Source → GitHub Actions.
4. Open Actions and wait for green checks.
5. Visit https://montrealai.github.io/skillos/

What works after upload

GitHub Actions will automatically:

1. verify the repository files;
2. run the SkillOS implementation tests;
3. run the full SkillOS loop;
4. generate dist/data/demo.json from the live demo;
5. generate dist/data/wealth_proof.json from the economic proof;
6. publish the website to https://montrealai.github.io/skillos/.

The website is deliberately static HTML/CSS/JS, so it is a good fit for GitHub Pages.

What the demo proves

The repo now proves two things.

First, it demonstrates the core SkillOS loop:

Work → Trace → Learn → Skill → Test → Approve → Release → Improve

A successful build produces a visible demo showing:

jobs captured → lesson discovered → candidate skill created → test passed → canary release approved

Second, it proves one concrete wealth-accumulation workflow:

sales follow-up email from call notes
→ every job creates a trace and lesson
→ every lesson creates a tested skill release
→ cost/job decreases
→ minutes/job decrease
→ quality score increases

Run it locally with:

python -m skillos.cli wealth-proof

The proof report is generated at docs/wealth_accumulation_proof.md.

Local run for technical users

No API keys are required.

python -m skillos.cli demo
python -m skillos.cli wealth-proof
python -m skillos.cli status
python -m skillos.cli serve

Then open:

http://127.0.0.1:8765

Repository contents

.github/workflows/pages.yml      Deploys the website
.github/workflows/tests.yml      Runs verification and tests
site/                            Public GitHub Pages website source
scripts/build_pages.py           Builds the static website, demo JSON, and wealth proof JSON
scripts/prove_wealth_loop.py      Proves one workflow gets cheaper, faster, and better
scripts/verify_repo.py           Fast repository sanity check
skillos/                         Python SkillOS reference implementation
tests/                           End-to-end and Pages build tests
docs/                            Architecture, roadmap, governance, troubleshooting
COPY_PASTE_GITHUB_ACTIONS/       Backup workflow files for GitHub web upload issues

Best first click

README_FIRST_GITHUB_WEB_USERS.md