Claude Code Confidence

Interactive beginner course

The terminal isn’t scary. It’s just a conversation with receipts.

Learn to use Claude Code safely, clearly, and confidently—even if you do not consider yourself a developer. You’ll practice the commands, make judgment calls, and finish with a repeatable workflow for real projects.

7 bite-sized modulesHands-on terminal labScenario practiceFinal score + certificate

🎯

What you’ll be able to do

Install and start Claude Code, give it project rules, choose safe permissions, guide complex work, verify changes, and troubleshoot common problems.

🧭

How this course works

Each module follows a simple loop: learn the idea, see it in action, make a decision, then prove you can use it.

🛡️

Your safety mindset

Start in Ask mode, use version control, review every diff, keep secrets private, and never let speed outrun judgment.

✨

No jargon gatekeeping

Every new term is translated into normal language. Open the glossary anytime if you need a quick refresher.

Module 1 • 10 minutes

Meet the terminal—without the drama

The terminal is a text-based way to tell your computer what to do. Claude Code lives there so it can inspect a whole project, suggest edits, run allowed commands, and show you the evidence.

🎯

By the end: You can explain what Claude Code does, choose the right installation path, and verify that it is ready.

Claude web chat

You bring snippets to the conversation. It answers based on what you paste or upload.

Claude Code

It works inside a project folder, can read the structure, propose file changes, and run approved commands.

Before you install: You need a supported account or API access, internet, a terminal, and a supported operating system. Use the native installer unless you specifically need npm/version pinning.

Choose your setup path

curl -fsSL https://claude.ai/install.sh | bash

Runs the recommended native installer. Read commands from trusted sources before pasting them into a terminal.

irm https://claude.ai/install.ps1 | iex

Run in PowerShell. Windows users may also work through WSL depending on their setup.

npm install -g @anthropic-ai/claude-code

Requires Node 18+. Avoid using sudo with npm; fix the user-writable prefix instead.

Install
Use the path above that matches your system.

Verify

claude --version claude doctor

Authenticate
Run claude, sign in through the browser, then return to the terminal.

Quick check

What is the safest default for someone brand new to Claude Code?

✓ Module complete. You have the setup map and the right safety starting point.

Module 2 • 10 minutes

Start a session and speak clearly

Claude Code is strongest when it is inside the correct project folder and your request names the outcome, location, limits, and proof of success.

🎯

By the end: You can navigate to a project, start Claude, and turn a vague request into an actionable prompt.

Practice terminal

Type cd my-project, then claude, then explain project.

The prompt recipe: O-L-L-P

Outcome

What should be true when the task is done?

Location

Which page, feature, folder, or file is in scope?

Limits

What must stay unchanged? What should it ask before doing?

Proof

Which tests, checks, or acceptance criteria prove success?

Prompt makeover

Which request gives Claude the clearest path?

Useful session commands: claude -p "task" runs one task and exits; claude -c continues a recent session; /resume reopens from inside the CLI; /clear starts fresh.

Quick check

Your prompt is producing edits in the wrong places. What should you add first?

✓ Module complete. You can start a session and give Claude a usable brief.

Module 3 • 10 minutes

Give Claude a project playbook

A CLAUDE.md file is the “how we work here” note Claude reads at session start. It reduces repeated explanations and prevents the assistant from inventing conventions.

🎯

By the end: You can draft a useful CLAUDE.md with context, stack, standards, and commands.

# My Project ## Project Context What this product does and who it serves. ## Stack - Frontend: React + Vite - Styling: CSS modules ## Code Standards - Keep components small and accessible - Do not change shared brand tokens without asking ## Commands - npm run dev — starts the preview - npm test — runs tests - npm run build — verifies production build

User level

~/.claude/CLAUDE.md
Preferences that should follow you across projects.

Project level

./CLAUDE.md
Shared rules that belong with the repository.

Directory level

./src/CLAUDE.md
More specific guidance for one area.

After an edit

Start a fresh session or use /clear when you need Claude to reread the instructions.

Build the better rule

A site uses teal and magenta brand tokens. Which CLAUDE.md instruction is most useful?

Quick check

Where should team-wide project conventions live?

✓ Module complete. Claude now has a dependable project playbook.

Module 4 • 12 minutes

Stay in control of every action

Permissions decide what Claude can do without stopping to ask. Convenience is useful—but only after you understand the action and its blast radius.

🎯

By the end: You can distinguish safe routine actions from actions that should stay gated.

Ask mode

Best for beginners and unfamiliar projects. Claude requests approval for writes and shell actions.

Allowlisted mode

Pre-approve narrow, repeatable actions such as reading or running one known test command.

Skip permissions

Disables approval prompts. Reserve for tightly controlled automation—not interactive learning or everyday work.

Version control

Work inside Git so you can inspect the diff, create checkpoints, and recover from mistakes.

Safety sorter • Click each action to cycle through the three categories.

Categories: ASK → ALLOWLIST → BLOCK. Choose what a cautious beginner should do.

{ "permissions": { "allow": [ "Read", "Glob", "Grep", "Bash(npm test*)" ] } }

Never paste secrets into a project rule or prompt. API keys belong in protected environment variables or secret managers. Treat remote-control links like passwords.

Quick check

What is wrong with allowlisting Bash(*)?

✓ Module complete. You can trade speed for convenience without giving away control.

Module 5 • 12 minutes

Use the workflow that prevents chaos

For anything bigger than a tiny edit, slow down before the first file changes. Planning makes assumptions visible; verification makes results trustworthy.

🎯

By the end: You can guide a complex task through plan, build, review, test, and commit.

Plan
Use /plan. Ask Claude to identify files, risks, assumptions, and acceptance criteria.

Approve the approach
Correct misunderstandings before they become edits.

Build in small checkpoints
Keep scope narrow and ask before risky changes.

Review the diff
Inspect exactly what changed. A confident explanation is not proof.

Run tests and checks
Ask Claude to test its work, then inspect the results.

Commit and reset context
Create a recoverable checkpoint. Use /compact or /clear when the session gets crowded.

Choose your next move

Claude says the checkout redesign is complete, but it changed 19 files and did not run tests. What should you do?

The golden rule: Plan describes intent. Diff shows edits. Tests provide evidence. You need all three for meaningful work.

Quick check

When is /compact most useful?

✓ Module complete. You have a repeatable workflow that keeps quality visible.

Module 6 • 11 minutes

Automate the boring parts—carefully

Hooks, worktrees, parallel sessions, plugins, and schedules can multiply output. They also multiply mistakes when the task is unclear or the guardrails are weak.

🎯

By the end: You can identify the right power feature for a job and explain the guardrail it needs.

Hooks

Run predictable actions at lifecycle moments—format after an edit, block a dangerous command, or test at task completion.

Worktrees

Give parallel sessions isolated Git workspaces so two tasks do not edit the same files.

Loops & schedules

Repeat local tasks with /loop or use supported cloud scheduling for work that continues while your laptop is offline.

Plugins, MCP & skills

Connect approved tools and reusable instructions. Add only what the project actually needs.

{ "hooks": { "PostToolUse": [{ "matcher": "Write|Edit", "hooks": [{ "type": "command", "command": "npx prettier --write \"$FILE_PATH\" 2>/dev/null || true" }] }] } }

Match the feature

You want one session to update documentation while another fixes tests, with no file collisions. What is the best starting point?

Quick check

Which hook is the best place to automatically run a formatter after Claude edits a file?

✓ Module complete. You can add leverage without removing the guardrails.

Module 7 • 10 minutes

Customize, troubleshoot, and go mobile

A useful setup fits the way you work. Customize what helps you stay oriented; use mobile access for monitoring and approvals, not squinting through major reviews.

🎯

By the end: You can personalize your experience, use remote control safely, and solve common issues.

`/statusline`

Show useful context such as model, folder, and context usage.

`/color`

Give concurrent sessions different visual identities.

`/voice`

Speak detailed prompts when talking is easier than typing.

Output style

Choose explanatory or learning-oriented responses when you want coaching.

Remote-control checklist

Start from your active terminal session
Use the current supported remote-control command shown in Claude Code.

Open the secure link or QR code on your phone
Treat that link like a password and never post or forward it.

Keep the computer and terminal session running
Remote control connects to the local process; closing it ends the session.

Save major review for a larger screen
Mobile is ideal for monitoring, quick instructions, and low-risk approvals.

Troubleshooting triage

Claude keeps ignoring the CLAUDE.md rule you just edited. What is the best first step?

Feature note: Claude Code commands, plan requirements, and mobile availability can change. Confirm subscription and version-specific details in Anthropic’s current documentation before rollout.

Quick check

Your remote session disappears when you close the terminal. Why?

✓ Module complete. You are ready to use Claude Code as a governed, flexible collaborator.

Final challenge • 10 questions

Prove you can run the workflow

Score 80% or higher to complete the course. You can retry; the goal is confident decisions, not perfect memory.

Baddie Academy AI • Certificate of Completion

This certifies that

Course Graduate

successfully completed the beginner course in safe, effective terminal-based AI collaboration.

Plan clearly • Protect the project • Verify the work