Free learning / 35-minute terminal lab

Your first command: know what will run

Separate the terminal, shell, CLI and coding agent. Run a real local helper and inspect how arguments, folders and output channels change its result.

Begin reading Download a working copy

No account needed. Use your own copy for writing and keep the original as a reference.

Engineering lab library / Run locally

Understand the command before the agent.

Run one local helper in two folders. Read its actual report, diagnostic and exit code before deciding whether it answered the task.

Synthetic records. No provider key or model inference. Runs locally with the standard library.

Start here / Node.js 24 · Offline exercise
First command and working directory
Run five actual process cases and explain the report, diagnostics and exit code.
Node.js 24 · Standard library · No model or account
Open walkthrough Download file

A colleague asks you to inspect one ticket from North workshop. You will use a small command-line program, then deliberately run it in the wrong folder. The point is to understand what the command actually did before accepting an agent's explanation or a green process result.

This original 35-minute exercise uses Node.js 24 and fictional files. There is no model, provider account, package installation or customer service. The interactive version is in the protected Academy lesson; the runnable exercise and all instructions here are free.

1. Separate the window from the program

A terminal displays text and delivers your keystrokes. A shell interprets command lines and starts programs. Bash, zsh and PowerShell are shells, with differences in syntax. A command-line interface (CLI) is the operations and options a program accepts through text arguments.

Node.js is a runtime: it executes JavaScript outside the browser. Our .mjs file is the program it runs. A coding agent is a separate application that may use a model to propose commands and inspect results. An ordinary terminal command does not automatically involve an agent.

Text example

You type in the terminal
        |
        v
The shell interprets the command
        |
        v
Node.js runs the JavaScript helper
        |
        +--> stdout: the report
        +--> stderr: the note or diagnostic
        +--> exit code: the process result

Think of a function call with an operation and inputs. The analogy has a limit: a command also has a working directory, an environment and system permissions. Those details matter when an FDE works across customer systems.

2. Prepare the small local workshop

← Back to lab library Download First command and working directory ↓

Download First command · Node.js into a new folder. Open your terminal in that folder. Check your runtime and run the reproducible comparisons:

Code example

node --version
node fde-first-command.mjs test
node fde-first-command.mjs compare
node fde-first-command.mjs init

Use Node.js 24. Eight checks exercise actual child processes, folder-dependent reads, argument handling, diagnostics, bounded input and preservation of an existing workshop. compare prints five cases from temporary synthetic files and removes only its own temporary directory.

init creates this new folder beside your download:

Text example

fde-first-command.mjs
cli-workshop/
  north/
    project brief.json   → North workshop, tickets N-1 and N-2
  south/
    project brief.json   → South workshop, tickets S-1 and S-2

If cli-workshop already exists, init refuses to overwrite it. Keep your work and use a different parent folder if you need a fresh workshop. inspect reads its chosen file; it does not edit it. Installing Node itself may need internet access; the exercise uses the installed standard library and makes no network requests.

3. Read the command before pressing Enter

In Bash, zsh or PowerShell:

Code example

cd cli-workshop/north
node ../../fde-first-command.mjs inspect --file "project brief.json" --limit 1

In Windows Command Prompt:

Code example

cd cli-workshop\north
node ..\..\fde-first-command.mjs inspect --file "project brief.json" --limit 1

Piece	What it does here
`node`	Runs the JavaScript runtime
`../../fde-first-command.mjs`	Names the script two folders above the current location
`inspect`	Selects the helper's operation
`--file`	Names an option
`"project brief.json"`	Groups a filename with a space into one argument
`--limit 1`	Supplies an option and its value

The grouping quotes are not part of the filename. The helper receives the filename as one string. Its --limit value also arrives as text; the helper checks that it is from 1 to 3 before converting it to a number.

The comparison runner passes argument arrays to a fixed Node executable with shell: false. It records actual process outputs, but it does not exercise a shell parser, PATH lookup or a coding agent. The displayed commands are the equivalent commands for a learner to type.

In an ordinary shell, PATH helps locate executable names such as node. Built-in commands such as cd may be handled by the shell itself. Use the installed shell's help for its exact rules. Do not treat every tool's flags as interchangeable.

4. Follow the working directory

The current working directory is the folder from which relative file paths start. It is not necessarily the folder containing the script.

After the North command, move to the neighboring folder and repeat it. In Bash, zsh or PowerShell:

Code example

cd ../south
node ../../fde-first-command.mjs inspect --file "project brief.json" --limit 1

In Windows Command Prompt:

Code example

cd ..\south
node ..\..\fde-first-command.mjs inspect --file "project brief.json" --limit 1

The arguments are unchanged. The report now names South workshop because that is where the relative filename was read. Both runs can exit 0. Only the North result answers the original task.

Change --limit 1 to --limit 2 and predict the ticket count. Then change it to --limit 0: the helper should refuse that value. Change one input at a time so the observation has a clear cause.

5. Inspect three separate results

stdout, standard output, contains this helper's JSON report. stderr, standard error, contains its note or diagnostic. The exit code is a separate numeric result when the program finishes.

Case	stdout	stderr	Exit code
North folder, ordinary arguments	North report	Training-fixture note	0
South folder, same arguments	South report	Training-fixture note	0
Missing filename	Empty	Input diagnostic	2
Misspelled `--limti`	Empty	Usage diagnostic	2
Limit outside 1–3	Empty	Usage diagnostic	2

A terminal may show both channels together. Text on stderr does not necessarily mean failure: the ordinary run writes an explanatory note there. This program defines 0 as completion and 2 as refused arguments or input. Other programs define their own nonzero results.

Check status immediately after the command, using the syntax for your shell:

Code example

# Bash or zsh
echo $?

Code example

# PowerShell: after a native executable such as node
$LASTEXITCODE

Code example

REM Windows Command Prompt
echo %ERRORLEVEL%

A later command can change the saved status. Do not type all three variants; choose the one for the shell you are using.

Try the refusal from either workshop folder:

Code example

node ../../fde-first-command.mjs inspect --file missing.json --limit 1

Use ..\..\fde-first-command.mjs in Command Prompt. Expect an empty report, an input diagnostic and exit code 2. The helper limits input to 8 KiB and does not include supplied file contents or raw exception paths in diagnostics. Its JSON report is local output from the file you explicitly selected.

6. Explain what a successful command does not prove

The South run is a successful process and an unsuccessful answer to the North task. Keep both facts. A model summarizing the output cannot change which folder the process used.

Before approving an agent's proposed command, identify the executable, working directory, arguments and expected effects. Afterwards inspect the actual report, diagnostic and process result. This exercise only reads fictional files; it establishes no permission to run a command in a customer environment and measures no agent behavior.

Keep a short command note:

What task was requested?
Which program and operation ran, and in which folder?
What arguments and effects were expected?
What appeared on stdout and stderr, and what was the exit code?
Did that observation answer the intended task?

For a transfer exercise, imagine an agent reports that tests passed but the command ran in a neighboring checkout. Preserve that output, identify the intended checkout and test command, and obtain evidence from the right place.

Continue with repository orientation in the Claude Code course. It uses the existing offline ticket-triage project and a real failing test.

Primary references and further learning

Checked 20 October 2026. The workshop, program and instructional examples above are original LockedIn Labs material.

Node.js: arguments, working directory and process output
Node.js 24.12: child processes and shell behavior
Node.js 24.12: parsing command-line options
MIT Missing Semester: Course overview and the shell — the creator's 2020 Unix-oriented lecture and exercises. Use current documentation for runtime installation and shell-specific differences.