Percerno Skillsbeta
← everything it can do · documents

Extract From Text PPTX

Pull structured values out of any file in ONE command, without reading the file into the conversation: e-mail addresses, phone numbers, web addresses, dates, money amounts, IBANs, Danish CVR numbers, CPR-shaped numbers (masked), invoice numbers, or any custom pattern, from PDF, Word, PowerPoint, Excel, text, Markdown, HTML or a whole folder; unique values with counts as a CSV. Use for: get every e-mail address out of this export, list all the invoice numbers in these PDFs, which phone numbers are in this document, find all amounts in this contract, check a file for CPR numbers before sending it. Triggers: extract, pull out, list all the, find every, e-mails in, phone numbers in, amounts in, invoice numbers, CVR, CPR check, regex.

1 uses this weeksource Percernolicence MIT
Quality gate: PASS · fewer tokens, same result · measured 2026-10-24
Same outcome with 0 % fewer tokens than reading the file into the model (1 requests).
1/1 requests answered with the skill doing the work, through the connector; direct run 3/3
Security scan: clean · 2026-10-24 · 2 files
no network calls, shell tricks, credential access or instruction tampering found in its files.

Try asking

Say it in your own words; the assistant finds this recipe by itself. Examples:

“Use the extract-from-text skill to …”
Try this skill from Claude, ChatGPT or a chat on your own page. Free plan, no credit card.Get started free

Details

When it is the right choice

Pull structured values out of any file in ONE command, without reading the file into the conversation: e-mail addresses, phone numbers, web addresses, dates, money amounts, IBANs, Danish CVR numbers, CPR-shaped numbers (masked), invoice numbers, or any custom pattern, from PDF, Word, PowerPoint, Excel, text, Markdown, HTML or a whole folder; unique values with counts as a CSV. Use for: get every e-mail address out of this export, list all the invoice numbers in these PDFs, which phone numbers are in this document, find all amounts in this contract, check a file for CPR numbers before sending it. Triggers: extract, pull out, list all the, find every, e-mails in, phone numbers in, amounts in, invoice numbers, CVR, CPR check, regex.

The full recipe the assistant reads

Extract values from files (one command)

python3 /skills/extract-from-text/scripts/extract.py IN OUT.csv --what <kinds> — IN is one file or a folder in /work/in; OUT gets columns type,value,count,files. The file text never has to pass through the model: the OK line summarises counts and the CSV holds the values.

Need Command
E-mail addresses extract.py /work/in/export.xlsx /work/out/emails.csv --what emails
Several kinds at once extract.py /work/in /work/out/found.csv --what emails,phones,urls
Money amounts and dates in a contract extract.py /work/in/contract.pdf /work/out/amounts.csv --what amounts,dates
Invoice numbers across scanned PDFs extract.py /work/in/invoices /work/out/invoices.csv --what invoices
Privacy check before sending extract.py /work/in/report.docx /work/out/pii.csv --what cpr,emails,phones (CPR values are masked ddmmyy-****)
Your own pattern extract.py /work/in/log.txt /work/out/ids.csv --regex "ORD-\d{6}" --name order
Show the first 15 in the OK line add --show 15

Kinds: emails, phones (Danish and international), urls, dates (dd.mm.yyyy, dd/mm/yyyy, yyyy-mm-dd, "24. oktober 2026", "24 October 2026"), amounts (kr., DKK, €, EUR, $, USD, "1.250,00"), ibans, cvr (8 digits labelled CVR), cpr (ddmmyy-xxxx), invoices (faktura/invoice no.). Text is taken from PDF (pdftotext), Word (pandoc), PowerPoint, Excel and plain files; scanned PDFs without a text layer need OCR first (the pdf-to-markdown skill).

Report the counts per kind from the OK line and give the CSV link; quote a few values only if asked.

Files in this skill (3)
  • SKILL.md
  • scripts/__pycache__/extract.cpython-311.pyc
  • scripts/extract.py
Where it comes from

Written by Percerno, MIT licence.