SkillsSkill for Claude

Move-out condition comparison

Give Claude the move-in and move-out walkthrough videos of your rental units and the parts on your move-in checklist. Claude prices the batch against your Vivu plan, uploads it to a private project, runs one search per part across every video, checks each returned window on a contact sheet, puts a move-in frame and a move-out frame of each part side by side, and writes only the differences both frames show. You add the finding and any deduction. The checklist says good or fair; whether the counter was replaced or has a burn mark is only in the footage, and the skill never decides what is damage or who pays.

Maintained by Vivu. Updated 2026-11-19.

Download

move-out-condition-comparison.zip

9 KB. Unzips to move-out-condition-comparison/SKILL.md. Upload the zip as it is in the Claude app, or unzip it into your skills folder for Claude Code.

SHA-256 b730369494542579d53ad13040cbf8f6a688e7cf882de52c3b619880b5ea8484

At a glance

What the Move-out condition comparison skill does, where it runs, what it needs, and when it asks
Looks forThe same kitchen counter and bathroom vanities in two walkthroughs of a unit shot months apart, often once furnished and once empty. Walkthroughs rarely narrate condition, so the part and its state are only in the picture. Every window is checked on a contact sheet against the source video before a frame goes into the document.
Runs onClaude Code on your computer (terminal or the Code tab of Claude Desktop), because it reads your local video files and runs ffmpeg. No downloads, so no residential IP needed, and nothing is scheduled.
Needs
  • The Vivu connector with write access, to create a private project, upload and search.
  • Both walkthrough videos of every unit as files on your computer, because every frame is cut from the local file.
  • ffmpeg and ffprobe, to measure the videos and cut contact sheets and side by side images.
  • A browser to open the Vivu upload page, or a browser tool that can attach local files.
  • Your move-in checklist or the list of parts to compare, because the parts drive the searches.
Your Vivu planIndex minutes equal the total length of the videos in the batch. Each part is one precise search across every video in the batch, and a precise search uses 5 credits in total, so adding units adds minutes but not credits. Claude shows the cost against your plan with vivu_get_usage and waits for your approval before uploading.
Asks you firstConfirming which file is each unit's move-in and move-out video, confirming the walkthroughs were recorded under the lease or your inspection procedure, uploading the batch and spending its index minutes, writing all comparison documents after you approve one sample unit, and any attachment of the documents outside the working folder.

The skill does its video work through the Vivu connector. If the connector is not in your Claude yet, add it first; the skill checks that it is connected before it does anything else.

Before you run it

  • Use only walkthroughs recorded under the lease or your inspection procedure; lived-in units show belongings and children's rooms, which the documents leave out.
  • The skill writes observations, not findings: it does not decide damage, normal wear and tear or deductions, which follow your state and city deposit rules.
  • Vivu's visual search is rated weak overall and does not say which bathroom a window shows; bathroom matches can be UNCERTAIN and a missing view means not found by search, not undamaged.
  • In our test run the vanity search returned 17 windows, 16 real and 1 false; the false one was a bathroom seen through a doorway, which is why every window is checked on its frames.
  • Walls, floors, windows and appliances are untested parts; check their first windows on contact sheets before trusting them.
  • Videos stay in your Vivu project until you delete them; the project is created private because the default is visible to your whole organization.
  • Claude in Chrome uploads at most 10 MB per call, so walkthrough videos usually go through your own browser or the Vivu web app.
  • The skill sends nothing; attaching the documents anywhere is your step, or Claude asks before each write.

Start it

Once the skill is installed, ask for the task in your own words. Naming the skill is the most reliable way to have Claude use it. For example:

Compare the move-in and move-out walkthrough videos in ~/Rentals/October for units 4B and 12, kitchen and bathrooms, and make a side by side comparison for each unit's deposit review.

In Claude Code you can also type /move-out-condition-comparison. Claude asks for anything the request leaves out, most important first.

What is inside

  1. When to use
  2. Working principles
  3. What you need before starting
  4. Inputs to collect
  5. Files and state
  6. Step 1: Check the Vivu connector and the setup
  7. Step 2: Pair the files, measure them and price the batch
  8. Step 3: Upload and index
  9. Step 4: Search each part across all files
  10. Step 5: Check every window on a contact sheet
  11. Step 6: Pair each part and put the frames side by side
  12. Step 7: Write the comparison documents
  13. Compliance
  14. Known failure modes

The full skill

This is move-out-condition-comparison/SKILL.md from the download, as Claude reads it: the frontmatter first, then the instructions.

---
name: move-out-condition-comparison
description: "Line up the same parts of a rental unit from its move-in and move-out walkthrough videos with Vivu, side by side, for a deposit deduction review. Use when a tenant moves out."
---

Move-out condition comparison with Vivu

This skill turns the move-in and move-out walkthrough videos of one or more rental units into one comparison document per unit. The property manager names the parts to compare from the move-in checklist (kitchen counter and sink, bathroom vanity, and so on). Claude measures the videos, prices them against the user's Vivu plan, uploads them to a private Vivu project, runs one precise search per part across all the files, sorts the returned windows into each unit's move-in and move-out column by file name, checks every window on a contact sheet, puts one move-in frame and one move-out frame of each part side by side, and writes only the differences those two frames show. The manager fills in the finding and any deduction.

The value is in what the checklist cannot hold. A move-in checklist says "good" or "fair" next to "kitchen counter"; it does not say whether there was a burn mark, which faucet was there, or whether the counter was replaced since. That is only in the footage, and walkthrough videos rarely narrate condition. Finding the same part in two videos shot months or years apart, often once furnished and once empty, is the slow part a manager skips. Vivu's search only points at candidate moments, and its windows often start or end in the next room, so every window is checked on a contact sheet before it reaches the document, a difference is written only when both frames show it, and the skill never decides what is damage, normal wear, or who pays.

When to use

Use when someone asks to compare move-in and move-out videos of a rental, to prepare evidence for security deposit deductions, to find the kitchen or the bathrooms in two walkthroughs of the same unit, or to line up a turnover inspection against the move-in record.

Not for these:

  1. One question about one video ("where does the move-out video show the oven?"): search Vivu directly and look at the frames.
  2. Deciding deductions or liability. Deposit rules differ by state and city, and normal wear and tear usually cannot be charged; the manager decides, and the skill writes observations only.
  3. Inspections narrated out loud, where the inspector says each finding: the inspection-findings-list skill (https://vivu.ai/skills/inspection-findings-list) lists spoken findings.
  4. Photos instead of videos: open the photos side by side; there is nothing to search.
  5. Identifying people, or listing a tenant's belongings: the skill never searches for people and leaves belongings out of the document.

Working principles

  1. Report measured numbers, not estimates. When a number is an estimate, say so.
  2. Nothing is verified until it has been checked against the source video. Vivu's windows are candidates; a frame enters the document only after Claude has looked at it, and the manager's finding is the decision.
  3. Stop and tell the user when a required capability or tool is missing. Do not guess around it.
  4. Ask the user before anything that is expensive to redo (the file list and its index minutes, before upload) and before writing all documents (one rendered unit first). The skill sends nothing; if the user asks Claude to attach the documents somewhere, ask again before each write.
  5. Write a difference only from the two frames, never from Vivu's reason text, which describes what was asked for. When the angle, the light or the resolution is too different to tell, write UNCERTAIN and say why.
  6. An empty cell does not prove the part was not filmed. Write "not found by search" and let the manager scrub the video.

What you need before starting

Check each item at the start of the run and tell the user plainly what is missing before doing anything else.

Requirement Why How to check
Vivu connector with write access create a private project, open its upload page, search vivu_get_account shows can_create_projects: true (tool names may carry a server prefix). A write call failing with "has not granted vivu.write" means the user reconnects Vivu and allows write access
Both walkthrough videos of every unit as video files on the user's computer Vivu indexes video files, and every frame is cut from the local file ffprobe -v error -show_entries format=duration -of csv=p=0 FILE prints a length for each file; download them from the phone, shared drive or property software first
A shell on that computer with ffmpeg and ffprobe measure the files, cut contact sheets and side by side frames ffmpeg -version, ffprobe -version
An upload path move the files into Vivu vivu_open_upload_page plus the user's own browser, or a browser tool that can attach local files
The move-in checklist, or the list of parts to compare the parts drive the searches ask for it; the default is the two parts in Step 4

This skill needs Claude Code on the user's computer (the terminal or the Code tab of Claude Desktop), because it reads local video files and runs ffmpeg. Claude on the web and cloud sessions cannot reach the files. It downloads nothing from the internet, so no residential IP is needed. Nothing recurs, so no scheduler is involved; run it once per batch of move-outs.

Inputs to collect

Ask for anything missing, most important first.

  1. VIDEO_DIR: the folder that holds the move-in and move-out videos. Required.
  2. Which file is which: for every unit, its move-in file and its move-out file. Required; Claude proposes the pairing from the file names and the user confirms it.
  3. The parts to compare. Default: kitchen_counter and bath_vanity from Step 4. Other checklist parts are added as candidates (see Step 4).
  4. BATCH_NAME: a short name for this batch (for example the month). It names the working folder and the Vivu project. Default: the name of VIDEO_DIR.
  5. Who reviews the documents. Default: the user.

Files and state

Keep everything in one working folder next to VIDEO_DIR:

moveout-BATCH_NAME/
  config.json      VIDEO_DIR, parts with query and maximum_results, Vivu project id
  units.csv        unit, visit (movein or moveout), source_file, duration_s, listed_name, video_id
  results/         search results without the result page link, one JSON per part
  sheets/          contact sheets per part and window
  pairs/           UNIT_PART.png side by side images
  UNIT_comparison.md   one per unit
  state.json       steps done, files uploaded, parts searched with job ids, window keys checked, pairs written

The commands in Steps 2 to 6 run from the folder that holds both VIDEO_DIR and moveout-BATCH_NAME/, so the relative paths in them resolve.

A rerun reads state.json first and skips finished steps. A file already marked uploaded is never uploaded again, a part with a saved result is not searched again, a window key (video_id, start_ms and part) that has a judgment is not checked again, and a pair already in pairs/ is not cut again. The next batch gets its own folder and its own Vivu project, so its searches do not return this batch's windows.

Step 1: Check the Vivu connector and the setup

Goal: confirm every requirement before spending anything.

  1. Call vivu_get_account. If the tool does not exist, tell the user to add the Vivu connector in Claude (https://mcp.vivu.ai/mcp) and stop. If the account does not show can_create_projects: true, or a later write call fails with "has not granted vivu.write", ask the user to reconnect Vivu and allow write access, then stop until they have.
  2. Run ffmpeg -version and ffprobe -version.
  3. List VIDEO_DIR and run the ffprobe length command on one file.

Done when vivu_get_account shows can_create_projects: true, ffmpeg and ffprobe print their versions, and ffprobe prints a length for a file in VIDEO_DIR.

Step 2: Pair the files, measure them and price the batch

Goal: a confirmed move-in and move-out pair for every unit, and an indexed set the user's plan can pay for, approved before upload.

  1. List VIDEO_DIR and propose units.csv: one row per file with unit, visit (movein or moveout) and source_file. Show it to the user and wait for an OK; a wrong pair makes every row of that unit's document wrong. A unit with only one video stays out of the batch, and the user hears which.
  2. Run the ffprobe length command on every file and fill duration_s. Add listed_name: the file name with each space, bracket, plus sign and percent sign replaced by an underscore, because that is how Vivu will list it.
  3. Call vivu_get_usage and show one table:
This batch Plan allowance Remaining this month
Index minutes sum of duration_s / 60 from vivu_get_usage from vivu_get_usage
Search credits 5 per part, 10 for the default two parts (estimate), 35 for seven parts from vivu_get_usage from vivu_get_usage

Plan facts: Free is $0 a month with 20 index minutes a month and 50 search credits a month; Premium is $30 a month with 180 index minutes and 500 search credits. A precise search uses 5 credits in total. Each search covers every file in the batch, so adding units adds index minutes but not credits. Six units with two six minute walkthroughs each are about 72 index minutes (estimate): more than Free covers in a month, inside Premium.

  1. If the batch does not fit, offer levers in this order: only the units whose deposit is being settled now, a second batch next month, and only then a larger plan. Never drop a file the user asked for without saying which.
  2. Confirm the Compliance items with the user.

The account used in our test run returns no allowance figures, so the comparison with a real plan and the user's approval were not exercised in our test run.

Done when the user approves units.csv and the batch's index minutes, and config.json records both.

Step 3: Upload and index

Goal: every file in units.csv indexed in a private Vivu project for this batch.

  1. Call vivu_list_projects and reuse a project named "Move-out BATCH_NAME" if one exists. Otherwise call vivu_create_project with that name and visibility "private". The default visibility is organization, which shows the project to everyone in the user's Vivu organization, and the inside of tenants' homes is sensitive.
  2. Call vivu_open_upload_page with the project ID immediately before uploading. The link expires in 180 seconds and works once, so never post or store it. Give it to the user to open in their own browser and select the files, or open it in a browser tool that can attach local files. Claude in Chrome accepts at most 10 MB per upload call and walkthrough videos are usually larger, so they normally go through the user's own browser or the Vivu web app. Never split or recompress a file to fit.
  3. Poll vivu_list_videos until every file shows ready. Match each Vivu file name to units.csv by listed_name and record its video_id there and in state.json.

In our test run the six files went through the same upload page with an automated browser, so opening the link in the user's own browser was not exercised in our test run. The files sat queued behind other indexing for several minutes and were all ready about 10 minutes after the upload started.

Done when every file in units.csv shows ready and has a video_id.

Step 4: Search each part across all files

Goal: a saved list of candidate windows for every part.

Field Mode maximum_results Status
kitchen_counter precise 40 tested
bath_vanity precise 40 tested
wall_by_room, floor_by_room, window_blinds, appliances precise 40 candidate, not exercised in our test run

The tested queries, word for word:

kitchen_counter: the kitchen counter with the sink and the stove or cooktop in view
bath_vanity: a bathroom vanity: the bathroom sink and its countertop, usually with a mirror or toilet nearby

The chain: one precise search per part finds windows in every file at once; file names put each window into its unit's move-in or move-out column; the contact sheet in Step 5 confirms the part and picks the frame; Step 6 compares the two frames. Every search is precise, because only precise returns a time window; fast returns whole files with an empty reason. Write each part as a room plus a fixed object in it ("the kitchen counter with the sink"), not as a surface that every room has. Walls, floors and windows exist in every room, so a search for "a wall" returns every room; if the user needs them, search them per room ("the bedroom wall beside the window"), and treat the results as untested until their first windows have been through Step 5. There is no second modality to switch to: walkthroughs rarely say anything about condition, so the comparison rests on the picture. Finding parts rests on Vivu's visual search, which our capability notes rate weak overall.

maximum_results is also the ceiling on how many windows come back. 40 leaves room for two or three bathrooms in each of several units; a search that returns exactly 40 was cut off. Then tell the user, mark the part cut off in state.json, and offer to split the batch.

Run each query with vivu_search_videos (project_id, query, mode, maximum_results). It returns a job ID. Call vivu_get_search_results until complete is true; each status call can wait up to 45 seconds, so a pending search is not a stalled one. Save each completed result as results/PART.json without its result_page_url field. Show the Vivu result page link in the live reply only; it expires after four hours, so it never goes into a saved file or a document.

Done when every part has a completed result saved in results/, and every part whose search returned exactly its maximum_results is marked cut off in state.json.

Step 5: Check every window on a contact sheet

Goal: every window judged "shows the part" or "does not show the part" from its frames.

For each result in results/PART.json, cut a contact sheet of the window from the local file at two frames a second:

mkdir -p moveout-BATCH_NAME/sheets/PART
ffmpeg -nostdin -v error -y -ss START -to END -i FILE -vf "fps=2,scale=240:-2,tile=6x6" moveout-BATCH_NAME/sheets/PART/rRANK_STEM_STARTMS_%02d.png

PART is the field, FILE the local source file (VIDEO_DIR/ followed by its source_file), STEM its name without .mp4, RANK the result_number written with two digits, STARTMS the result's start_ms, and START and END the result's start_ms and end_ms divided by 1000. Each sheet holds 36 tiles covering 18 seconds; tile k on sheet n (both counted from 0, left to right and top to bottom) is at START + 18 * n + k / 2 seconds. Keep -nostdin when the command runs inside a loop that reads the result list.

Look at every sheet. Do not judge a window from its first, middle and last frame alone: a walkthrough window often starts in the hallway and ends in the next room, and the part is in between. A window shows the part when at least one tile shows it clearly enough to compare (for a vanity: the sink and its counter, not only a toilet or a mirror). A part seen from the next room through a doorway does not count. Record the window key, the judgment and the best tile's second in state.json.

Done when every result in results/ has at least one sheet in sheets/ and a judgment in state.json.

Step 6: Pair each part and put the frames side by side

Goal: one side by side image per unit and part, with the differences the two frames show.

  1. For each unit and part, take the windows judged "shows the part" from its move-in file and its move-out file. A unit has one kitchen, but often two or three bathrooms, and the search does not say which bathroom a window belongs to. Match bathrooms between the two videos by fixtures (tub color, tile, vanity size, mirror and lights) and by the order of the walk, name them bathroom 1, 2 and 3, and mark a match UNCERTAIN when the fixtures do not agree. A part with no window in one video gets "not found by search" in that column.
  2. Pick the tile in each window where the part is largest and the angle is closest to the other video, and cut both frames into one image:
mkdir -p moveout-BATCH_NAME/pairs
ffmpeg -nostdin -v error -y -ss IN_SECONDS -i IN_FILE -ss OUT_SECONDS -i OUT_FILE -filter_complex "[0:v]scale=-2:360[a];[1:v]scale=-2:360[b];[a][b]hstack=inputs=2" -frames:v 1 moveout-BATCH_NAME/pairs/UNIT_PART.png

IN_FILE and OUT_FILE are the unit's move-in and move-out source files, IN_SECONDS and OUT_SECONDS the chosen tile times, and PART carries the bathroom number when there is more than one (bath_vanity_2). Scaling both sides to the same height keeps a lower resolution video from shrinking its half. Open the image; if a side shows the mirror or the shower instead of the part, pick another tile and cut it again.

  1. Write visible_difference from the image only: what is in one frame and not the other (a different counter surface, a missing fixture, a stain, a hole), in plain words, labeled observed. Write "no difference visible" when the shared area looks the same, and say what limits it (lower resolution, glare, a different angle). Leave belongings and people out: a furnished move-in frame shows the tenant's things, and the comparison is about the unit.

Done when every unit and part has a row with either a side by side image or "not found by search", and each row has a visible_difference.

Step 7: Write the comparison documents

Goal: one document per unit that the manager can attach to the deposit paperwork after adding the findings.

Show the user one rendered unit and the field mapping, and wait for an OK before writing the rest:

| part | movein_mmss | moveout_mmss | side_by_side | visible_difference (observed) | manager_finding | deduction |
|---|---|---|---|---|---|---|
| kitchen_counter | 01:22 | 01:10 | pairs/U1_kitchen_counter.png | Counter and backsplash are a plain light surface in move-in and a speckled stone-look surface in move-out. Cabinets, white range, white microwave and the wallpaper border look the same. | | |

The row above is from our test run; the user's rows carry their own unit names and times.

Column Source If unavailable
part config.json, with the bathroom number from Step 6 none
movein_mmss, moveout_mmss the chosen tile times, written MM:SS "not found by search"
side_by_side pairs/ from Step 6 none; the row says "not found by search"
visible_difference Claude's reading of the side by side image (inferred), labeled observed "UNCERTAIN" with the reason
manager_finding, deduction left empty for the manager empty

Head each document with the two file names and one line saying that the differences are observations from frames, not findings. List the checklist parts that were not searched, so nobody reads a missing row as "no damage". The document uses the user's own file names and times, never a Vivu result page link, so it stays usable after the link expires.

Then tell the user, per unit, which parts have both frames, which were not found by search, and which bathroom matches are UNCERTAIN, and for each part how many windows came back and how many Step 5 rejected. Say plainly that a search misses some views and that the manager should scrub the video for any part that matters to a deduction.

The documents stay in the working folder. Attaching them to a tenant ledger, an email or property software is the user's step; if the user asks Claude to do it, name the destination, show the rendered document and wait for a yes before each write. In our test run the documents were rendered from the dry run's pairs; showing a sample to a user and handing the documents over were not exercised in our test run.

Worked example from our test run

The test corpus was public: one property management company's listing walkthroughs of homes it re-rented, three addresses listed twice each, with the earlier listing standing in for move-in and the later for move-out. Six videos (29.9 minutes) with the audio removed, handheld, no people on camera. One unit was furnished and lived in at the first listing and empty at the second; one move-out video was lower resolution than its move-in video. Kitchens, vanities and a brick fireplace were marked by hand on contact sheets before any search, and a hot tub search ran as a control, with no hot tub in any file.

  1. kitchen_counter: 7 returned, 7 real and 0 false, and 3 missed of the 10 kitchen visits marked by hand. Every unit's first kitchen visit came back in both videos; the misses were short second looks. Windows were around 15 seconds.
  2. bath_vanity: 17 returned, 16 real and 1 false, 0 missed. The false one was a bathroom seen from the hallway through a doorway. Windows were around 10.5 seconds, and in several of them the first, middle and last frames showed a mirror, a shower or the next room; only the contact sheet showed the vanity.
  3. A part only one unit has, a brick fireplace: 4 returned, 4 real, 2 missed, all from that unit's move-in and move-out videos and none from the other units. One window ended a few seconds before the camera looked at the fireplace again.
  4. The hot tub control returned 0 windows. No wording was changed on any search.
  5. Indexing the 29.9 minutes took about 10 minutes from the start of the upload to all six files ready, most of it with the files queued behind other indexing. Each precise search took between 4 and 5 minutes to complete.

The side by side images found a real change in one unit (the kitchen counter and backsplash were replaced between the listings) and no visible difference in the kitchen and bathrooms of another. Matching bathrooms was the weak link: in the two units with three bathrooms, some pairs stayed UNCERTAIN because the tile or the vanity length did not agree between the videos.

Done when every unit in units.csv has a UNIT_comparison.md with one row per part, the user approved the sample unit, and state.json marks the batch done.

Compliance

  1. Use only walkthroughs recorded under the lease or the company's inspection procedure, with the tenant informed as that procedure requires. Have the user confirm this before Step 3.
  2. Walkthroughs of lived-in units show belongings, photos, mail and children's rooms. The documents describe the unit only; frames with a person in them are replaced by another tile. The skill does no face or identity recognition and never searches for people. If footage is meant to show a person or a child, do not upload it.
  3. The skill does not decide damage, normal wear and tear or deductions. Deposit rules (what can be charged, deadlines, itemized statements) differ by state and city; the manager applies them.
  4. The videos stay in the user's Vivu project until the user deletes them. Create the project as private; the default is visible to the whole Vivu organization. Delete videos or the project only when the user asks, and confirm first.
  5. The skill writes only local files; attaching the documents anywhere goes through the approval in Step 7.

Known failure modes

Symptom Cause Fix
(observed) a bathroom seen only through a doorway came back as a vanity: 17 returned, 16 real and 1 false; the reason said "visible through the doorway" the search matches the room, not a view of the part close enough to compare judge from the sheet; a part seen from the next room does not count
(observed) the first, middle and last frames of a bathroom window showed a mirror, a shower or the next room and no sink windows cover the walk into and out of the room cut the full contact sheet before rejecting a window
(observed) short second looks at a kitchen were missed: 7 returned, 7 real and 3 missed one window per visit, and brief returns to a room are the easiest to miss the first visit is usually enough; scrub the video when a missing view matters to a deduction
(observed) a fireplace window ended at 02:08 and a second look a few seconds later was not returned one window does not cover every look at a part when the chosen frame is poor, scrub a little past the window end
(observed) a unit with three bathrooms returned windows that do not say which bathroom they show the search finds the part, not its room number match bathrooms by fixtures and walk order; mark UNCERTAIN when the fixtures disagree
(observed) the first side by side pick showed the mirror or the shower instead of the sink the tile with the most bathroom in it is not always the one with the vanity open every pair image and re-cut from another tile
(observed) the six files sat queued for several minutes before indexing ("Videos 0 indexed · 6 indexing") the account's indexing queue was busy keep polling vivu_list_videos; queued is not an error
(observed) a furnished move-in frame and an empty move-out frame of the same kitchen the unit was lived in when the first video was shot compare only what both frames show and leave belongings out
upload page asks to sign in or shows an error the one time upload link expires after 180 seconds request a new link right before opening it
a file is rejected by the browser upload tool the tool accepts at most 10 MB per call the user adds the file in their own browser or the Vivu web app
"has not granted vivu.write" Vivu connected read only the user reconnects Vivu with write access
a result's file name does not match units.csv Vivu replaced a space, bracket, plus or percent sign in the name match on listed_name in units.csv
a search returns exactly its maximum_results the batch holds more windows than one search returns mark the part cut off, say so, and offer to split the batch
walls, floors or windows return every room those surfaces exist in every room search them per room, and treat them as untested until checked on sheets

Connect Vivu, then add the skill.

The skill runs through the Vivu connector. Add it to Claude from the connector directory, then install the skill.