Tutor playbook

How to learn with Theo Marchand

A practical, profile-specific playbook for learning Terminal Coding Agents and Repository Automation, Claude Code, Codex CLI, Copilot CLI, shell workflows, permissions, approval modes, sandboxing, MCP...

Updated July 29, 2026 6 min read Build, inspect, test, and explain

View Theo's profile Compare tutor fit

Theo Marchand Terminal coding agents and automation tutor

On this page

Best fit

Is Theo Marchand right for your goal?

Developers who want terminal-first Claude Code, Codex CLI, Copilot CLI, MCP, and repository automation to remain observable and reversible.

Learning focus: Terminal Coding Agents and Repository Automation, Claude Code, Codex CLI, Copilot CLI, shell workflows, permissions, approval modes, sandboxing, MCP servers, hooks, non-interactive tasks, Git discipline, environment setup, secrets, logs, and failure recovery
Best level: Developers, DevOps learners, maintainers, platform teams, technical founders, and experienced students using terminal-first coding agents
Lesson format: CLI session walkthroughs, permission maps, sandbox diagrams, command reviews, MCP design exercises, automation scripts, Git recovery drills, environment checklists, and incident-style retrospectives
Languages: French, English

Theo is most comfortable when a terminal agent's trust boundary can be drawn on one page. He examines permissions, commands, secrets, Git state, tools, and recovery paths with quiet precision before allowing automation to become ambitious.

Calm Analytical Security-minded Dryly humorous Operational

Strong starting points

Terminal Coding Agents and Repository Automation
Claude Code and Codex CLI
Permissions, approvals, and sandboxing
MCP servers and hooks
Git safety, secrets, logs, and recovery

Before lesson one

Plan a focused first session

Specific evidence gives Theo a better starting point than a broad request to teach the whole subject. Use this four-part setup.

Arrive with evidence
Bring a code sample, error, command, diagram, dataset, requirement, or system behavior. A real sample gives Theo something concrete to diagnose.
Define one result
Aim for one working technical artifact plus a clear explanation of why it behaves that way. Put that result in the Plan tab before expanding the lesson.
Attempt before the model
Show what you currently think or can do. Ask for a hint or question before requesting the completed answer.
Leave with retrieval
Explain the lesson back, save the hardest point as a review card, and schedule the smallest useful follow-up.

Choose the right lesson mode

Text chat

Pasting errors, comparing approaches, reviewing code carefully, and preserving exact commands or explanations.

Paste the exact material and state the feedback format you want.

Voice call

Thinking through architecture, explaining a bug aloud, interviewing, and checking conceptual understanding.

Think aloud and ask Theo to pause after each correction or question.

Classroom

Shared code, architecture diagrams, debugging traces, documentation, and testable worked examples.

Share each workspace explicitly so the tutor can see the latest version.

Repeatable value

Use Theo's lesson rhythm

A good session should produce something you can attempt, inspect, and revisit. This profile is designed around the following rhythm.

Start: Identify the terminal agent, repository, operating environment, allowed tools, approval expectations, secrets, and rollback strategy.
Work: Inspect state, define the command boundary, run or simulate one agent task, review tool calls and diffs, test the result, and record recovery steps.
Continue: Permission matrices, sandbox designs, safe command reviews, MCP tool schemas, Git recovery drills, and bounded automation tasks.

Collaborative classroom

Use each classroom tool with a purpose

The whiteboard opens as the main lesson surface. Workspaces are shared deliberately: use Show tutor or Update tutor after changing the board, Notebook, Document, Code Lab, or SVG Studio so Theo sees the current version.

Whiteboard

Classroom

Trace state, data, control flow, dependencies, and assumptions before changing code or infrastructure.

Best move: Draw or place the first version yourself, then use Show tutor or Update tutor so Theo can respond to the current board.

Code Lab

Classroom

Use Code Lab for the smallest reproducible example, keep line numbers visible, and ask for tests or checkpoints before a full solution.

Best move: Keep one task active, ask for the smallest useful hint, and make a second attempt before requesting a complete model.

Practice

Classroom

Predict the result first, run or inspect the example, explain the difference, and then solve one nearby variation.

Best move: Ask Theo to adjust difficulty after each attempt and explain the exact cue that should transfer to the next example.

Notebook and Document

Classroom

Keep a debugging log with symptoms, hypotheses, evidence, the smallest useful change, and the lesson to reuse later. Write the requirement, architecture decision, API contract, or explanation beside the implementation so intent stays testable.

Best move: Keep your wording and decisions visible. Share the workspace with the tutor before asking for a revision or structured feedback.

Review Cards and Transcript

Classroom

Save commands, patterns, failure modes, and explain-it-back questions as review cards after the code works.

Best move: At the end, turn only the highest-value ideas and repeated mistakes into cards, then use the transcript to recover evidence or phrasing.

SVG Studio

Classroom

Turn architecture, data flow, state, and dependencies into a clean diagram that stays beside the code.

Best move: Ask Theo to label boundaries and failure points, then explain the diagram without looking at the implementation.

Theo's methods

Profile-specific teaching tools

These methods come directly from this tutor profile. The surface label shows where to make the result visible during a classroom lesson.

Code Lab

Shows shell commands, configuration, hooks, MCP examples, automation scripts, and tests with line numbers and syntax highlighting.

Try it with Terminal Coding Agents and Repository Automation in Code Lab, make one attempt yourself, then ask Theo to correct only what blocks the next step.

Code Lab

Whiteboard

Diagrams trust boundaries, local and remote environments, permissions, secrets, network access, Git state, and rollback paths.

Try it with Terminal Coding Agents and Repository Automation in Code Lab, make one attempt yourself, then ask Theo to correct only what blocks the next step.

Code Lab

Practice

Runs command-approval scenarios, prompt-injection defenses, tool-schema reviews, recovery drills, and automation design exercises.

Try it with Terminal Coding Agents and Repository Automation in Code Lab, make one attempt yourself, then ask Theo to correct only what blocks the next step.

Notebook

Document

Creates runbooks, permission policies, environment setup notes, secret-handling checklists, and incident retrospectives.

Try it with Terminal Coding Agents and Repository Automation in Notebook, make one attempt yourself, then ask Theo to correct only what blocks the next step.

Ready to use

Prompts that fit this tutor

These prompts use Theo Marchand's actual subjects, lesson format, and adaptive classroom lab. Replace the topic with your own material when needed.

Identify the terminal agent, repository, operating environment, allowed tools, approval expectations, secrets, and rollback strategy.
I want to improve Terminal Coding Agents and Repository Automation. Use CLI session walkthroughs, permission maps, sandbox diagrams, command reviews, MCP design exercises, automation scripts, Git recovery drills, environment checklists, and incident-style retrospectives. Check what I can already do, let me attempt something, and give one correction at a time.
Open Code Lab for Claude Code and Codex CLI. Keep the task small, make me explain my choices, and finish with three review cards and one next-session goal.

Progress evidence

Know whether the lessons are working

Do not measure progress only by how clear the explanation felt. Look for changes in what you can retrieve, decide, produce, or explain without support.

Predicts behavior before running the example
Finds the failing boundary with fewer hints
Explains tradeoffs instead of naming tools only
Builds a nearby variation without copying the model

Responsible use

Use Theo as a tutor, not an authority

Requires authorization, least privilege, protected secrets, explicit destructive-action review, and human validation before external or production effects.

Theo Marchand is a fictional AI tutor profile for terminal coding agents and automation tutor education.

Put the guide into practice

Start one focused lesson with Theo

Bring one real example, choose one result, and keep your first attempt visible. You can review the full profile before opening chat or voice.

View tutor profile Review prompts