From the video

Anthropic dropped mods for Claude Code. Here is the desk from the reel, yours to install.

Terminal Desk puts a stats bar above your prompt and a /desk pane with six panels: what Claude is assuming, what it left undone, where your context went, the prompt cache, subagents, and cost. Download the zip, unzip it into your Claude Code skills folder, and start a new session. While Claude works it fills in blanks on your behalf, and once your cache goes cold the next message resends the whole conversation. The desk shows you both while you can still act. The PDF explains every figure and what to do about it.

The guide covers what a mod is, the install on Windows and Mac, and what every panel and bar figure means. The mod download is in step one.

The download is free. What we sell is at the bottom of the page, with prices.

Rewatch on Instagram
  1. 01Download the mod and unzip it into your skills folder
  2. 02Reload plugins and open /desk
  3. 03Send back work Claude put off, and reject an assumption that is wrong
The mod and the guide
Free
Steps
4
Panels in the /desk pane
5
Unzipped into your skills folder
1 zip

Want this built for your business, or have an idea that needs AI?

Work with us

What you will have when you finish

The desk from the reel, running in your own terminal.

  • A stats bar above your prompt: working or ready, tokens, cost, context used, cache hot or cold, tool calls per minute, subagents, turns and errors
  • A list of work Claude put off, each with a Do it now button that drafts the instruction for you to send
  • The last five judgment calls Claude reported, each with a Reject button that drafts a correction you finish and send yourself
  • A view of what is filling your context window and how much room is left
  • A countdown on your prompt cache, so you know whether it will still be hot when you come back

Four steps

A mod is code that runs inside Claude Code with your permissions, so read the specifics below before you install one from anybody, us included. Mods require Claude Code v2.1.287 or later.

Claude Code v2.1.287 or later, running in a terminal. Check with claude --version. Tested on the Windows terminal, Claude Code v2.1.289.

  1. 01

    Download the mod, then have Claude Code unzip it into your skills folder. A plugin folder placed there loads in every session.

    Download the Terminal Desk mod

    Paste into Claude Code

    I downloaded terminal-desk-mod.zip to my Downloads folder. Unzip it into my personal Claude Code skills folder so this file exists: C:\Users\<me>\.claude\skills\terminal-desk\.claude-plugin\plugin.json. Create the skills folder if it does not exist. The zip already holds one top-level folder named terminal-desk, so do not nest it inside a second terminal-desk folder.
    
    Then confirm that plugin.json exists at that exact path and list the files you unzipped. Do not change any of them.

    Your part: approve the file commands when Claude Code asks. Anthropic's warning covers every mod, this one included: "A mod is code that runs with your permissions. It can read and write your files, start processes, and make network requests. Install mods only from authors and marketplaces you trust." The code is one file, hooks/register.tsx. Read it first if you want to.

  2. 02

    Load it and open the desk. In a session that is already open, type /reload-plugins. Otherwise the mod loads the next time you start Claude Code. The bar appears above the prompt, and /desk opens the pane.

    Type in Claude Code

    /desk

    Your part: check the bar is there. No bar? Run claude --version, because mods require v2.1.287 or later. Then check the folder did not end up one level too deep, at skills\terminal-desk\terminal-desk. To try the mod for one session without installing it, start Claude Code with claude --plugin-dir and the path to the unzipped folder.

  3. 03

    Read the panels, top to bottom. What Claude is assuming lists the last five judgment calls Claude reported. Left undone lists work Claude put off, taken from its own words, from markers it wrote into files, and from a second model that reads its final report. Where your context went shows what fills the window and the room left. Prompt cache says hot or cold, with the time left. Subagents lists each one with its status. Cost and tokens totals the session.

    Optional: give the panels something to show

    Look through this project and tell me which three files you would read first to understand it, and why. Do not change anything.

    Your part: watch the numbers move, then act on them. The context figure on the bar is green under 60, amber from 60 and red from 80. On Pro or Max the cache lasts one hour after your last request, and each request resets it. On an API key or a cloud provider, type /desk cache 5m. The PDF covers what to do about each figure.

  4. 04

    Act on what the desk caught. In Left undone, each entry has a lettered Do it now button. Press its letter, or click the button, and the mod drafts "You left this undone" in your prompt box with the entry quoted, ready to send. In What Claude is assuming, each open entry has a numbered Reject button that drafts the start of a correction: the assumption quoted back, then "Instead:". Nothing is sent until you press Enter.

    a Do it now

    Your part: decide. Send back work you asked for and Claude put off. Reject when Claude settled on a scope you did not ask for, took the wrong reading of your request, or chose to keep or delete something you would have chosen differently. Leave an entry alone when it is right. The mod never sends, rejects or corrects anything for you.

To remove the mod, delete the terminal-desk folder from your skills folder. /desk hide hides the bar and /desk show brings it back.

The specifics

What it costs
The mod and the guide are free. On Pro or Max, the dollar figure estimates what the same tokens would cost at API list prices.
Tested on
The Windows terminal, on Claude Code v2.1.289. Mac and the Desktop app are untested.
What it reads
Session usage, tool call names, subagent activity, and the text Claude writes, to spot put-off work. It writes no files.
What it adds to a session
A system prompt section of roughly 200 tokens, one tool, and one small Claude Haiku request after a turn with five or more tool calls.
Works where
Claude Code v2.1.287 or later. The bar and pane draw in the terminal CLI, editor terminals included, and the Desktop app's Code tab.
Managed plans
Team, Enterprise and managed machines: Claude Code stops a mod you install from changing the system prompt, so the assumptions panel may stay empty.

Questions people ask

  • Is it safe to install a mod?

    A mod runs with your permissions, like any program you install. Anthropic says to install mods only from authors you trust. This one is a single code file in the zip. Read it first, or ask Claude Code to explain it.

  • Does it send my code anywhere?

    Only to Claude, on your own plan. After a long turn the mod sends your request and the final answer to a small Claude model to check for skipped work. Type /desk check off to stop that. Nothing goes to any other service.

  • Does the dollar figure mean I am being charged?

    On Pro or Max, no. Usage is included in your plan, and the figure estimates what the same tokens would cost at API list prices. On an API key, treat it as an estimate and check your provider's usage page for the bill.

  • How do I get rid of it?

    Delete the terminal-desk folder from your Claude Code skills folder and start a new session. To keep the mod and lose the bar, type /desk hide.

Stuck on a step? Ask me.

This guide is also a lesson in my community, with its own questions thread for the step that stopped you. Joining is free. Paid members also get my ask me anything call every Tuesday at noon ET, and the full recordings of each setup as I make them.

The free tools people open most

See every free tool
  • Guide

    Reverse Engineer Anything with Claude Code

    Claude Code · Windows and Mac

    Get the guide
  • Skill and prompts for Claude Code

    Dynamic workflows kit

    Claude Code · 5 min of your time

    Download the kit
  • Claude Code and Codex plugin. Windows.

    Reel Studio

    Claude Code or Codex · Windows · 15 to 25 min

    Get it free

That was the free part

The file is yours. Nothing above asked for an email.

If it earned a minute of your time, the paid part is one scroll down.

The paid part

Here is what we sell.

Everything above this line is free and stays free. Below it is what we sell. The prices are the prices, and the two subscriptions carry 30 days money back on the first month.

Line chart of views by the hour for our five newest Reels, on a log scale. Two climb fast in the first hours. Three level off in the low thousands.
A real dashboard Claude built from our own account through the Analyzer. Last updated 15 October.

Mica Social Media Analyzer

Price: $29/ month

Your Instagram, answered by the ChatGPT or Claude you already use. Connect the account once, then ask in plain words what worked and what to post next.

You post on Instagram and you are done guessing from the Insights tab.

  • Which Reels held attention, ranked, without opening Insights
  • Readings at 1, 24 and 72 hours after every post, so you see what happened while it was moving
  • Next captions written in the style of your most saved posts
  • Read only. It never posts or changes anything

7 days free with a card on file. About a dollar a day after that.

Start 7 days free

30 days money back on the first month. Email guru@graniteai.co and we refund it.

Nine Instagram posts from an insurance agency in New Hampshire, one a day from 8 to 16 October. Fall scenes with headlines about crash calls, home rebuild costs, renewals and a customer review.
The Instagram feeds of two Mica Social customers, captured 16 October. Mica wrote, designed and published every post shown, one a day.

Mica Social

Price: $49/ month

An AI that writes, designs, checks and publishes a post to your Facebook and Instagram every day. You send a photo when you have one. It does the rest.

You know the feed should be alive every day and it is not.

  • A finished post every day, with a photo made for it. Nothing to shoot
  • Timed to your local weather, season and county numbers
  • Five checks before anything publishes. No approvals, no calendar, no logins
  • Every photo you send becomes its own extra post

An agency running a daily feed bills $500 to $1,500 a month. 3 days free with a card on file.

See Mica Social

30 days money back on the first month. Email guru@graniteai.co and we refund it.

The readiness audit, done for you

Price: $349One week

Send us access to your repo. We run the nine-lane review, change nothing, and hand back the report with a 30 minute call on the three gaps to close first.

You have a product you charge for, or want to, and you would rather read the verdict than run the agents.

  • The report at docs/commercial-readiness.md: a verdict, a scorecard, every gap with the file to open and how to verify it
  • Legal, security, billing, support, marketing, operations and the rest, checked at once
  • A 30 minute call on the three gaps to close first
  • Nothing in your repo changed. The workflow prompt is yours to run afterward

The review is the one in the video. On our own product it found 87 gaps. The overnight run that followed was 148 commits and one pull request.

If the review finds no gap worth your time, you pay nothing.

The guided first run

Price: $9960 minutes

We run the review on your repo together, on a call, and you leave with the report in your repo and the three gaps to close first.

You have the kit and a repo, and you would rather do the first run with someone who has done it.

  • The review running on your machine, in your Claude Code
  • The verdict read together, and the three gaps picked
  • The workflow prompt set up so you can run it that night
  • The call recording

New. You would be one of the first. The run itself is the one in the video: 190 agents, 148 commits, one pull request.

If the review will not run on your setup during the call, there is no charge.

GraniteAI builds and runs AI systems for small and midsized businesses, out of Manchester, New Hampshire.