← Back to library

Diagnose a down machine from its history, sensors and manual

Anthropic's Claude for Manufacturing plugin, released November 12, 2026, includes a skill for the moment a machine stops. It opens with a lockout reminder and any safety flag, ranks possible causes with the prior work orders behind each, filters out boilerplate close-out notes and leaves every ticket update to the technician. We explain how to use it.

  • OutcomeUse the breakdown-diagnosis skill from Anthropic's Claude for Manufacturing plugin to put any safety hazard first, read the sensor trace before the stoppage, search the asset's work orders and technician notes, check the manufacturer's troubleshooting guide, and get ranked likely causes with the quickest test to tell them apart plus a draft preventive maintenance change.
  • SetupAn afternoon+
  • LevelAdvanced
  • Creditanthropics
View original
Works with
ClaudeClaude CoworkClaude Code

Published Nov 14, 2026Updated Nov 14, 2026AI dry run on sample input: passed Nov 14, 2026Source verified Nov 14, 2026

When should you use it?

  • A machine has stopped and the technician needs a starting point faster than digging through old tickets
  • The same asset keeps failing and you want to know whether the maintenance schedule is the problem
  • You want a ranked list of the assets causing the most downtime across the plant

Not for: Replacing your lockout or safety procedures; the skill reminds you of them but does not assess hazards; Writing changes straight into the maintenance system; every output is a draft; Plants with no maintenance history, manuals or photos to give it

What do you need, and what do you get?

You need

  • The asset, the time it stopped, the symptom in the technician's words and any fault codes
  • Work order history with close-out notes, from a CMMS connector or an export
  • A sensor export covering the period before the failure, if the machine is instrumented
  • Photos of the failed area and the manufacturer's manual, where available

You get

  • A safety line first, with a flag and reason when the evidence shows a hazard
  • Ranked candidate causes, each with evidence, a first check and the expected fix
  • The prior work orders cited for each candidate, with boilerplate notes labeled
  • A draft change to the preventive maintenance interval or procedure for the planner
  • A downtime record ready for the next production report

How does it work, step by step?

  1. Install and onboardAdd Claude for Manufacturing from the Claude plugin directory and complete the plant setup so the skill knows your sites, systems and rates.
  2. Report the stoppageTell it which machine is down, when, what the technician sees and any codes on the display. Add photos if you have them.
  3. Give it the historyConnect the maintenance system or attach an export of work orders, plus any sensor data and the manual.
  4. Run the first checkRead the safety line, then do the cheapest test it names to separate the top candidates before opening anything up.
  5. Close the loopConfirm the cause, paste the drafted close-out note into the ticket yourself and pass the maintenance interval proposal to the planner.

The prompt

Our adapted version, written to work in Claude, Claude Cowork and Claude Code. Values in lime are placeholders.

Fill in the blanks (optional; the copied prompt uses your values)

Prompt7 variables
Machine {asset} stopped at {time}. The technician reports: {symptom}. Fault codes: {codes}.
Start with our lockout reminder and tell me whether anything in the evidence suggests a safety hazard, and why.
Then use {work_orders}, {sensor_export} and {manual} to rank the likely causes. For each one, give the evidence, the quickest check that would rule it in or out, and the expected repair.
Cite the work order numbers you relied on, ignore close-out text that appears word for word on unrelated machines, and draft a change to the maintenance interval for the planner. Do not update any ticket.

The original skill is published by anthropics on GitHub (anthropics/claude-plugins-official). Open it there to install or import it; the prompt above is our own adaptation.

What can go wrong?

  • Skipping the safety line because the list of causes looks more useful. Lock out the machine first, every time.
  • Trusting a nightly export for a breakdown that happened twenty minutes ago. Add the most recent tickets or describe recent work out loud.
  • Reading too much into terse notes such as 'fixed' or 'OK now'. The skill reports how usable the notes were; take that seriously.
  • Anthropic labels the plugin a reference implementation it does not actively maintain, so trial it on a past breakdown first.

Our take

Maintenance teams rarely lack data; they lack the time to read it while a line is down and a supervisor is asking for an estimate. This skill pulls together the four places a technician would look if they had an hour, then returns a short ranked list with the quickest test to narrow it. Two design choices stand out. Safety comes first in a fixed position, so nobody reaches for a wrench before reading about stored energy or a bypassed interlock. And the template check is clever: when the same close-out text shows up on machines from different makers, it is treated as a default someone clicked rather than evidence. We also like that the loop back to preventive maintenance is part of the job rather than an afterthought, because recurring failures usually trace to a schedule nobody revisited. Cautions: it writes nothing to your maintenance system, the plugin is labeled a reference implementation, and the output is only as fresh as your export. In our dry run on a fictional lathe stoppage, it led with a stored-energy flag, ranked two causes and suggested a five-minute check to split them.

Where the time goes back: Most repair time on a stopped machine goes into working out what is wrong. A ranked shortlist with the prior tickets attached and one quick test to run cuts that search down to minutes.

FAQ

Does it need sensor data?
No. Most machines are not instrumented, and the skill treats that as the normal case, working from history, notes, photos and the manual.
What does it do with copy-paste close-out notes?
It groups identical notes across the fleet. Text that appears on machines of different makes is shown as a template, not cited as evidence for this machine.
Who made it and what license does it use?
Anthropic published it in the anthropics/claude-plugins-official repository on GitHub under the Apache 2.0 license.

Source and credit

This playbook is based on work by anthropics. We summarize it and link to it and write our own adapted prompt.

Original
breakdown-diagnosis skill (anthropics/claude-plugins-official, Claude for Manufacturing)
Creator
anthropics
Published on
GitHub (anthropics/claude-plugins-official)
Original date
Nov 12, 2026
License
Apache-2.0
Originally for
Claude
AI dry run
Nov 14, 2026: passed. Dry-run tested by our AI test agent on sample input. Safety first, template note excluded, correct top cause, cheapest check named, nothing written back.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails.
Verified
Nov 14, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.

The original is licensed under Apache-2.0 by anthropics. Modified: our adapted prompt and summary are our own wording; the original skill {"was unchanged at the source when we snapshotted it (live source later removed)" if src_removed(s) else "is unchanged at the source link"}.

Are you the creator? Claim or correct this page.

More in Operations & Admin

All 203 Operations & Admin playbooks →

Related playbooks

Operations & AdminSkill✓ AI dry run

Compare design options with a weighted score and stress test

Use the trade-study-report skill from Anthropic's Claude for Manufacturing plugin to turn a filled-in options matrix and test results into a ranked, sourced comparison, set aside any option that breaks a hard requirement, vary the weights to see whether the winner changes, and draft a decision memo an engineer reviews and signs.

Credit: anthropics

Under an hour
Operations & AdminSkill✓ AI dry run

Connect your agent to a system with no ready connector

Use the custom-connector skill from Anthropic's Claude for Manufacturing plugin to name the exact system and one data flow, check for an existing vendor or community MCP server before building anything, settle on the highest workable access route down to a scheduled export, record why, keep credentials out of the chat and explain how the chosen route is likely to break.

Credit: anthropics

An afternoon+
Operations & AdminRoutine✓ AI dry run

Get the plant report ready before the morning meeting

Schedule the morning-plant-pulse skill from Anthropic's Claude for Manufacturing plugin to combine the overnight exports from planning, production and maintenance, compute output against plan, OEE and its three parts, downtime ranked by minutes and by count and the cost of losses, then refresh one operations dashboard per site with 14- and 30-day trends, all without changing any business system.

Credit: anthropics

An afternoon+
Operations & AdminSkill✓ AI dry run

Rank likely root causes and draft the 8D from plant data

Use the root-cause-capa skill from Anthropic's Claude for Manufacturing plugin to line up sensor, inspection, maintenance and lot records on one checked timeline, run control-chart and clustering checks in code, rank several candidate cause chains with evidence for and against each step, name the cheapest test that separates them and draft an 8D that leaves verification and closure for people to fill in.

Credit: anthropics

An afternoon+