Plant Floor

Let AI connect the records. Keep the evidence and decisions with your team.

Get the free starter ↓
Explore the step-by-step guide & full report
See the workflow in actionGUIDED EXAMPLE · NO LIVE AI
01 / 04
IG
YOUThe assignment

Review this notification backlog. Find possible duplicates and missing information. Give me a report with the evidence.

▤backlog.csv24 notifications · 6 equipment tags
NOTIFICATION MAP6 equipment tags · schematic view
P-204P-204CV-12CV-12AC-03AC-03TK-01TK-01PK-08PK-08AHU-02AHU-02WN-1041↗WN-1058↗WN-1082↗WN-1103↗WN-1027↗WN-1063↗WN-1094↗WN-1112↗WN-1035↗WN-1071↗WN-1097↗WN-1120↗WN-1045↗WN-1088↗WN-1107↗WN-1124↗WN-1050↗WN-1091↗WN-1115↗WN-1129↗WN-1060↗WN-1099↗WN-1118↗WN-1132↗
01 / THE INPUT

Your CMMS export.

Notification IDs, descriptions and whatever context you already have.

→
02 / THE ASSISTANT

Your coding agent.

Reads the supplied instructions and carries the work through to a checked output.

→
03 / THE OUTPUT

Your review report.

A portable HTML workspace with editable recommendations and bulk review.

TAKE THE WHOLE WORKFLOW

Don’t start
from a blank prompt.

The starter includes the instructions, data checks, example backlog and report design. Your AI handles the first pass. You make it fit your plant.

Download the free starter ↓

MIT licensed · Node.js 22+ · your own AI access
No hosted model or API key setup in the starter.

Explore the finished example report ↗
  1. 01

    Download and unzip.

    The starter is a complete folder. Open example-report.html to see the output immediately.

  2. 02

    Open the folder in Codex.

    Add your export to input/. Give the agent access to this folder so it can read the data and create the report.

  3. 03

    Give it this assignment.

    The agent handles the files, checks and report generation. You don’t need to carry JSON between tools.

YOUR STARTING PROMPT

Set up and run this Backlog Cleaner. Follow AGENTS.md. Check the prerequisites, use the CSV in input/ if there is one, and take it all the way through preparation, analysis, validation, and the HTML review report. Do the file handling yourself. Keep every source record, flag uncertainty, and leave all recommendations for my review. If there is no CSV yet, show me the included example first. Give me the finished report and only ask questions the records cannot answer.

What’s inside the starter?
  • AGENTS.md + CLAUDE.md — the complete agent workflow
  • backlog.mjs — prepare, validate and generate
  • lib/ — editable validation and report code
  • examples/ — fictional data and prepared analysis
  • docs/index.html — an example ready for GitHub Pages
  • LICENSE — reuse and adapt under MIT
MAKE IT YOUR OWN

The structure is supplied.
The next version is yours.

Ask your agent to add your site’s vocabulary, change the review rules, or adapt the report to your planning meeting. Keep the evidence attached to every recommendation.

PREFER TO TRY IT IN THE BROWSER?

Keep your hands on the steps.

The browser workbench supports CSV import and a manual handoff to your AI.

Open the browser workbench →