When should you use it?
- Production, quality and maintenance numbers live in separate exports that never quite agree
- Someone asks why totals from the shop floor and the ERP do not match
- You want reports or quotes built from one checked dataset instead of raw files
Not for: Live sensor traces or high-frequency machine data; the skill does not carry them; Writing corrections back to the ERP automatically; it proposes fixes and waits for approval; Businesses with a single system of record, where there is nothing to join
What do you need, and what do you get?
You need
- The folder where your MES, quality, maintenance and planning exports land
- Optional ERP access (the skill's examples are NetSuite and Infor M3) and a maintenance system connector
- A date window, or the latest export drop, and someone who can approve proposed fixes
You get
- A spreadsheet dataset joined on part, work order, sales order, lot and asset, keeping original spellings
- A data quality report with freshness per source, rows in versus rows joined and mismatches by class
- A proposed correction for each mismatch class, with how many records it would fix
- A reconciled figures block so every count in the report adds up
How does it work, step by step?
- Install the pluginAdd Claude for Manufacturing from the Claude plugin directory and complete its plant setup so it knows your export folder.
- Request the datasetName the period you need, such as the previous calendar month, or let it work from the latest export drop.
- Read the freshness bannerBefore any figures, check each source's date coverage. If windows only partly overlap, choose whether to narrow or wait for new exports.
- Review mismatch classesGo through each class, its example rows and the fix it proposes, and approve only the fixes you understand.
- Feed downstream workPoint your reports and quotes at the consolidated dataset rather than the raw files.
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)
Combine {export_folder} with ERP records for {date_window} into one dataset. Before joining, tell me each source's last refresh and the dates it covers; if they do not line up, stop and give me the options. Match on part, work order, sales order, lot and asset, keep the original spellings, and log every cleanup rule you apply. Group every record that fails to match by cause, propose a fix with how many rows it would resolve, and change nothing until I approve each group.
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?
- Joining stale exports. If one file is weeks old, the mismatches are manufactured; let the freshness check stop the run.
- Giving a blanket yes. Approve each mismatch class after seeing its rule and examples, not all at once.
- Leaving the plant scope unset. Without it, item counts can include parts from other sites or subsidiaries.
- Anthropic labels the plugin a reference implementation it does not actively maintain, so have your operations or IT lead review the setup.
Our take
Plenty of AI reporting demos assume clean, connected data, but most small and mid-sized plants run on scheduled exports dropped into a shared folder. This skill starts from that reality and treats file-only as the normal case. Its most useful habit is checking freshness before joining anything: if two sources cover different weeks, it says so and offers two named choices instead of quietly producing a dataset full of false mismatches. We also like that it never stops at a count. Every unmatched record lands in a named class with a proposed correction and the number of rows that correction would fix, and nothing is applied without approval. Keeping the original spelling of every part number next to the normalized one is a small detail that will save floor staff real frustration. The trade-off is setup effort and a manufacturing-specific schema, so it fits plants and job shops far better than general businesses. In our dry run on fictional MES and ERP exports, it caught a two-week freshness gap, grouped revision-suffix and leading-zero mismatches with proposed fixes and waited for approval.
Where the time goes back: Reconciling exports from several plant systems by hand can take a planner or analyst hours every week, and the same key mismatches come back each time. A repeatable join with named mismatch classes and proposed fixes turns that into a short review.
FAQ
Do I need an ERP connector?
Will it change records in my systems?
Who made it and what license does it use?
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
- data-consolidate 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 17, 2026: passed. Dry-run tested by our AI test agent on sample input. Freshness before figures, every mismatch explained, approval per class; no live ERP connection or file output in the dry run.
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 17, 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.