Bots

Listing Command Center Cowork Agent

What This Is

The Listing Command Center is a Cowork agent that turns your listing’s source files into the marketing assets you need to launch and sell it. You drop the MLS report, photos, and comps into a folder. You ask for what you need in plain language. The agent reads what’s there and produces the asset.

Six outputs you can request:

  • Listing description — MLS Public Remarks written to be found by AI search engines like ChatGPT, Perplexity, and Google AI Overviews. Noun-dense, feature-rich, compliant.
  • Photo ordering — renames your photo files into proper MLS display sequence: front exterior first, logical walk-through flow, outdoor and garage last.
  • Photo captions — one MLS-ready caption per photo, in display order.
  • Deal of the Week email — a plain-text email that reads like a personal note, not a marketing blast. Two variants per run so you can pick the one that fits.
  • Just-listed HTML email — a branded campaign email, built in your colors and fonts, ready to paste into your ESP.
  • Pricing analysis — a full CMA with a sales-comparison adjustment grid, an appraised value, a marketable value, and a list price recommendation that bridges the two.

You drop the files. You ask. You review and use what comes back. The agent reads your source files before it does anything, so every output is grounded in the actual listing — not guesses.

Fair Housing and RESPA guardrails are baked in. The skills avoid school quality ratings, demographic framing, steering language, and lender recommendations. You should still review every output before it goes to your MLS, your database, or a client — AI guardrails reduce risk, but they don’t replace your professional judgment.


How to Use It

Setup takes about 20–30 minutes the first time. You’ll download a task folder, run an intake conversation, install your skills, and paste a task instruction into Cowork. After that, every new listing is just a few minutes of drop-the-files-and-ask.

You’ll need: the Claude Desktop app (Mac or Windows, any paid plan), and about 15 minutes to answer questions about your brand and your market.


STEP 1: DOWNLOAD YOUR TASK FOLDER

Download the zip file below and unzip it. The folder you get — listing-command-center-task-folder — is your task folder for the Listing Command Center. It contains the Claude.md system prompt, your agent profile template, a references folder, and seven skill folders you’ll use during setup.

Contains: Claude.md (system prompt), agent-profile-template.md (personalized during intake), a references folder, and seven skills — listing-remarks-writer, listing-photo-organizer, listing-photo-captioner, listing-html-email, dotw-email, email-subject-line-writer, and mls-data-analyzer. You’ll install the skills to your Claude skill library in Step 4 and delete the skills folder from your task folder once they’re installed.


STEP 2: MOVE YOUR TASK FOLDER WHERE YOU WANT IT

Move the unzipped listing-command-center-task-folder wherever you want it to live on your computer — Desktop, Documents, Dropbox, OneDrive, iCloud Drive, anywhere that works for you. This is the folder you’ll point Claude Cowork at when you run the task.

One note specific to this agent: your task folder will also be the home for every listing you work on. If you already have a “Listings” folder where you store property files, you can rename this folder to match or merge the contents into your existing folder. Either way works — Cowork just needs to know which folder to read.

If you need a template, you can download a sample folder structure: Click here.


STEP 3: RUN THE INTAKE PROMPT

Open a regular Claude chat — not Cowork. Before you paste anything, attach four files from your task folder: agent-profile-template.md, and the SKILL.md files from the listing-html-email, dotw-email, and mls-data-analyzer folders. These are the files Claude will personalize for you during the conversation.

Then copy and paste the intake prompt below into the chat. Claude will ask you about six bundles of questions covering who you are, your market, your brand, your voice, your email setup, and your local market knowledge — voice-dictate your answers if you can, it goes fast. When you’re done, Claude gives you four things:

  • A populated agent-profile.md — download this. You’ll place it in your task folder in Step 5.
  • A personalized listing-html-email skill — download this. You’ll install it in Step 4.
  • A personalized dotw-email skill — download this. You’ll install it in Step 4.
  • A personalized mls-data-analyzer skill — download this. You’ll install it in Step 4.

Here’s the intake prompt — copy and paste it into Claude:

Prompt
# Listing Command Center — Intake Prompt
You are setting up a personalized Listing Command Center for a real estate agent. The member has downloaded a task folder from the AiM library containing:
- **Claude.md** — the orchestrator system prompt (lives in their task folder)
- **agent-profile-template.md** — the blank profile you will populate (attached to this conversation)
- Seven skill folders. Three are shell skills you will personalize: **listing-html-email**, **dotw-email**, **mls-data-analyzer** (their SKILL.md files are attached). The other four install as-is.
**Four files should be attached to this conversation:** `agent-profile-template.md` and the `SKILL.md` files from `listing-html-email`, `dotw-email`, and `mls-data-analyzer`. If any are missing, tell the member which ones you need before proceeding.
Your job is to:
1. Interview the member in six conversational bundles
2. Populate agent-profile.md from their answers
3. Personalize the three shell skills from the attached SKILL.md files
4. Deliver everything as downloadable files
---
## How This Conversation Works
Ask the questions in six bundles. Bundle related questions into one natural message rather than 30 separate asks. Wait for answers before moving on. Be conversational and brief.
**Offer to help when the member doesn't know an answer.** If they don't know their hex codes or fonts, offer to inspect their website. If they skip something with a safe default, use it and tell them what you chose.
---
## Bundle 1: Who You Are
Name, title, brokerage, brokerage mailing address (for the RESPA email footer), license number(s), phone, email, website, scheduling link (optional), and whether they're solo or a team.
If they have a website, offer to pull contact details and gaps from it.
## Bundle 2: Your Market
Primary market area (city, state, county, primary ZIPs), 3–5 primary neighborhoods, and any additional licensed states.
## Bundle 3: Your Brand Visual Identity
Three brand colors (primary, secondary, accent) as hex codes, heading and body fonts, and four email design preferences: corner style (sharp/soft/rounded/pill), button shape (square/rounded/pill), density (compact/standard/airy), header treatment (solid/gradient/image).
If they don't know hex codes or fonts, offer to inspect their website. If they say "you decide," default to: soft corners, pill button, standard density, solid header — and tell them.
## Bundle 4: Your Voice
3–5 voice descriptors, standard sign-off, always-include language, and never-use language. Pre-load these never-use defaults and show them: never use "actually," never use all-caps for emphasis. Optional: 1–3 short writing samples for voice matching.
## Bundle 5: Your Email Setup
ESP (Mailchimp, ActiveCampaign, Constant Contact, HubSpot, Klaviyo, ConvertKit, Follow Up Boss, SendGrid, Flodesk, or other). Image URLs — logo, brokerage logo, headshot, Equal Housing logo — noting the ESP media library is the recommended host. Default CTA text for listing emails. DOTW default audience, recurring framing language ("Deal of the Week" or their variation), variant preference (both / question-led / numbered-reasons), and signature style. Home valuation tool URL, home search portal URL, and any other standing lead magnets.
## Bundle 6: Your Market Knowledge
Primary MLS, Public Remarks character limit (default 1,000 if unknown), any MLS-specific compliance flags, primary property types listed, and photo display style preference (standard / hero-shot lead / AI decides).
Then the four market-context questions — the most important fields in the intake. Push for specifics:
1. What do buyers in your market pay a premium for?
2. How do you define a comparable property?
3. What kills a deal in your market?
4. Which 2–3 data points do you always show sellers first when advising on price?
---
## After All Six Bundles
Confirm a brief summary. Then produce:
### Deliverable 1: agent-profile.md
Populate the attached template with their answers. Apply these defaults if skipped: body text #2D323C, background #F7F7F9, soft corners, pill button, standard density, solid header, DOTW both variants, first-name signature, MLS limit 1,000, standard photo display. Look up the member's ESP first-name merge tag syntax from the listing-html-email matrix and write it into the profile.
### Deliverable 2: Personalized Shell Skills
Personalize the three attached SKILL.md files by replacing their bracketed fields with the member's values:
- **listing-html-email** — identity, brand colors, fonts, design tokens, image URLs, ESP, defaults
- **dotw-email** — recurring framing, default audience, variant, signature, first name, voice notes, always-include / never-use phrases
- **mls-data-analyzer** — local market context (value drivers, comp criteria, deal-breakers, seller presentation lead)
Do not modify skill structure, output formats, or compliance rules — only personalize the user-specific fields. Remove every bracket. No placeholder text in the final files.
### How to Deliver
Save agent-profile.md and the three personalized skills as downloadable .md files. Then tell the member:
"Here's your populated agent profile and three personalized skills. Save all four. The library resource page walks you through installing your skills, deleting the skills folder from the task folder, placing your agent profile in the task folder, and pointing Cowork at it. You're almost done."
---
## Final Rules
- Save all deliverables as downloadable .md files.
- Personalize only user-specific fields in the shell skills. Leave structure, output formats, and compliance rules untouched.
- No brackets, no template syntax, no placeholder text in any final file.
- Keep instructions conversational and directive.
- If any of the four required files isn't attached, ask for it before proceeding.

STEP 4: INSTALL YOUR SKILLS

For each skill, open Claude Desktop → Customize → Skills → click + → Upload a new skill, drag and drop the file, save.

Install all seven. For the three you personalized in Step 3 — listing-html-email, dotw-email, mls-data-analyzer — install the personalized versions Claude gave you, not the templates from the task folder. For the other four — listing-remarks-writer, listing-photo-organizer, listing-photo-captioner, email-subject-line-writer — drag the SKILL.md file directly from the matching folder inside your task folder.

One exception: listing-remarks-writer includes a reference folder with an additional markdown file. You can’t install it by dragging just the SKILL.md — you need to zip the entire skill folder first. Right-click the folder, choose Compress, then drag the resulting zip into the skill upload modal.

Make sure each skill is toggled on. Once all seven are installed, you can delete the skills/ folder from your task folder. The skills live in your Claude skill library now and don’t need to stay in the task folder. Your task folder should contain only Claude.md, your agent profile (added in the next step), and the references/ folder.


STEP 5: ADD YOUR AGENT PROFILE

Inside your task folder, place the populated agent-profile.md file Claude generated for you in Step 3. It goes at the root of the task folder, alongside Claude.md. You can also delete the original agent-profile-template.md if it’s still there — you won’t need it again unless you re-run the intake.

Your task folder should now look like this:

listing-command-center-task-folder/
├── Claude.md
├── agent-profile.md
└── references/
    └── listing-folder-schema.md

STEP 6: SET UP THE COWORK TASK

Cowork runs on the Claude desktop app — it works by accessing files on your local machine, which is what makes it different from a regular cloud-based chat. Open the Claude desktop app (Mac or Windows) and switch to Cowork mode. Then:

  1. Start a new task
  2. Select Sonnet 4.6 as your model. This task is built to run on Sonnet — it handles the listing analysis, CMA work, and email writing well and keeps your usage manageable. If you choose a more powerful model like Opus, the output quality may improve slightly but each run will consume significantly more of your usage allowance.
  3. When Cowork asks you to select a working folder, choose your listing-command-center-task-folder from wherever you moved it in Step 2
  4. For the task instructions, paste this:
Follow the instructions in Claude.md in the task folder root. At the start of every session, scan the contents of the task folder and the relevant listing subfolder before producing any output, so you know what source files you're working with.

This task is on-demand — there’s no schedule. You open the chat for this task whenever you have a listing to work on.


STEP 7: RUN IT ONCE MANUALLY

Before you can run anything, you need to drop a listing’s source files into the task folder. Create a subfolder inside your task folder — name it whatever makes sense, the address works well: 123-main-street, 14-oak-avenue, 1402-elm-court.

Drop the source files for that listing into the subfolder. The Command Center looks for these:

  • MLS report — PDF export from your MLS (the full listing report). Required for most outputs.
  • Photos — .jpg, .png, or .heic files in the subfolder or a photos/ subfolder. Required for photo ordering and captions.
  • Tax record / sales history — optional, useful for CMA context.
  • Comps — CSV or XLSX of comparable properties. Required for pricing analysis. Best practice: pull a broad export — all statuses (active, pending, closed, expired, withdrawn), full year, broader area than just the immediate subdivision. The agent sorts what’s relevant.
  • Prior listing remarks — optional, if the property was listed before or you have a draft. Used as voice reference, never copied.

You don’t need every file for every command. The listing description runs on just the MLS report. The HTML email runs on the MLS report plus image URLs you provide in chat. Add files as you have them.

Now open the Cowork chat for your Listing Command Center task and ask for what you need in plain language:

Write the listing description for 123 Main.

The Command Center reads your agent profile, finds the listing subfolder, inventories the source files, extracts the property facts, researches nearby amenities, asks you about anything ambiguous, calls the right skill, and writes the output to that listing’s output/ subfolder.

You can request one output at a time, or several at once:

Generate everything for 123 Main — order the photos, write the description, caption the photos, build the HTML email, and run the pricing analysis.

It runs them in efficient order, pausing only when it needs input from you — like hosted image URLs for the HTML email.


STEP 8: WHAT YOU GET BACK

Output files land in the listing’s output/ subfolder. Each is a polished draft you review, edit, and use.

  • Listing description → listing-remarks.md — paste into your MLS Public Remarks field
  • Photo ordering → photos renamed in place, plus a report showing the changes
  • Photo captions → photo-captions.md — paste into your MLS photo captions field
  • DOTW email → dotw-email.md — two variants with subject lines and preheaders
  • HTML email → just-listed-email.html — copy the HTML and paste into your ESP’s code editor
  • Pricing analysis → cma-seller-facing.md (the narrative for your seller) and cma-internal-memo.md (your internal pricing rationale)

Source files stay where you put them. The Command Center only writes to references/ (its working memory) and output/ (your deliverables). The one exception is photo ordering, which renames photo files in place but never deletes them.


STEP 9 (OPTIONAL): SET UP YOUR NEXT LISTING

Every new listing is a few minutes of setup. The pattern repeats:

  1. Create a new subfolder inside your task folder, named after the listing
  2. Drop in the source files — MLS report, photos, comps
  3. Open the Cowork chat for your Listing Command Center task and ask for what you need

You don’t need a new Cowork task for each listing — keep using the same one. The Command Center handles multiple listing subfolders inside the same task folder, and figures out which one you’re asking about from your message.

If you want to queue a request while you’re away from your computer, use the Dispatch feature in the Claude mobile app. As long as your computer is on with Cowork running, you can trigger a request from your phone — useful when you’re at a listing presentation or driving and want to kick off a draft before you get back to your desk.


USAGE AND COST

Everything runs inside your Claude subscription — no outside APIs, no separate billing. Each command draws from your usage allowance, the same pool you use for regular Claude chats.

Most commands are light. The pricing analysis is the heaviest — it reads the comps, builds the adjustment grid, and writes two documents — but it’s still a single on-demand run, not a recurring job. If you’re watching your usage closely, run one output at a time rather than “generate everything,” and choose Sonnet 4.6 as your model. Sonnet handles every output in this agent well and keeps usage manageable. A more powerful model like Opus may sharpen the writing slightly but will consume more of your allowance per run.


Tips

  • The HTML email needs hosted image URLs. The Command Center doesn’t host images — you do. Upload your listing photos to your ESP’s media library, copy the public URLs, and paste them when the agent asks. Avoid Google Drive and Dropbox links — they often break in email clients.
  • Update your agent profile whenever something changes. New brokerage, refreshed brand, shifted market focus — edit agent-profile.md directly, or re-run the intake. The agent adapts to whatever’s in the profile.
  • Each listing’s working memory is cached. Property facts, amenity research, and the file inventory are written to references/ inside each listing subfolder. The second time you run a command for the same listing, it’s much faster — no need to re-extract data or re-do research.
  • Review before you publish. Every output is a draft, never auto-published. Read it. Check the local details and the numbers. Make sure the voice feels like you. You’re the agent of record — the output is your responsibility before it goes to your MLS, your database, or a client.