Training

Short lessons for the RobOS tools that plan work, run coding agents, and control what agents send and run. Each lesson uses commands and screens that exist in the repository and says where a feature is limited.

Lessons

Track Lesson You will Needs
Plan and run agent tasks Plan a feature in Task Planner Turn a template form or a description into an epic with tasks and create the issues. A GitHub or Jira task server, gh
  Check impact with the kgraph CLI Find what depends on a component and validate a graph change before an agent starts. Node.js 20 or later, a checkout
  Run an agent on a task Start Claude Code from Task Implementer on an issue and review the branch it pushes. A task server, claude, gh
  Choose and customize agent personas Pick the role prompt an agent gets and add your own. Task Implementer or Task Planner
Agent harness and security Configure the prompt security guard Redact or block secrets, card numbers and injection phrases in prompts. Node.js 20 or later
  Run commands in an ephemeral agent profile Give an agent a throwaway home directory, and know what that does and does not isolate. Linux, python3, Electron

Where the rest is documented

These lessons stop where another page already covers the topic.

Topic Page
Creating a project from a template New app wizard
Reviewing an agent’s branch, the quiz and evidence stages, signing off PR Review Theater
What the knowledge graph stores and how packages are laid out Knowledge graph
Which models and agent tiers RobOS uses for which job Agent tiers
Installing RobOS and opening the first app Get started

Running the apps

Open an app from the App Launcher, or run it from a checkout with Electron:

electron packages/task-planner

Replace task-planner with the package folder of the app. Every lesson names the apps it uses and links to their app pages.


Table of contents