Turn a PDF table into a usable Excel sheet

2. Prepare the skills

Open each guide for the upstream source, setup requirements, and a smaller first exercise. Installation depends on your agent and its available tools.

1. 1

   Prepare the two skills

   Open the PDF and XLSX skill pages below. Use their upstream installation instructions and confirm your agent can read local files and create a workbook.

2. 2

   Attach the PDF

   Download invoice-lines.pdf and attach it or place it in your agent workspace. The CSV is an answer key for you, not an extraction input.

3. 3

   Extract and preserve

   Ask the PDF skill for the six rows. Keep each item ID as text and record its page. Stop and label uncertainty if a cell cannot be read.

4. 4

   Build the workbook

   Use the XLSX skill to keep quantities and amounts numeric, calculate totals with formulas, and compare the sum with the printed subtotal.

5. 5

   Check before reusing

   Open the workbook, compare it with the PDF and CSV answer key, and verify the six acceptance checks below.

3. Run the first task

Attach the sample inputs in your agent, then send this prompt.

Use the PDF and XLSX skills on the attached invoice-lines.pdf. Create invoice-review.xlsx with sheets named Line items and Summary. Extract all six item rows into item_id, description, quantity, unit_price_usd, line_total_usd and source_page. Keep item_id as text with leading zeros. Use formulas for calculated line totals and the subtotal. Compare the sum with the printed PDF subtotal and show the difference in Summary. Recalculate and check formula errors. Also write reconciliation.txt with the row count and any unreadable values; never guess missing data. Return both files and the checks you actually ran.

4. Check the files you get back

Use these specifications to check your own result. Any recorded example outputs are listed separately below.

invoice-review.xlsx

Line items sheet with the six source rows and source-page references; Summary sheet with a formula total and reconciliation difference.

reconciliation.txt

Row count, extracted subtotal, printed subtotal, and any ambiguous cells that require review.

Acceptance checklist

• The Line items sheet contains exactly six data rows, with IDs 001, 002, 003, 004, 005 and 006.
• Each row records source_page = 1 and the currency is USD throughout.
• Line totals are 450.00, 400.00, 80.00, 125.00, 95.00 and 100.00.
• The formula subtotal is 1,250.00 and the difference from the printed subtotal is 0.00.
• Changing a quantity updates its line total and the summary after recalculation.
• No formula errors, guessed cells, dropped rows, or lost leading-zero IDs remain.

Limits of this exercise

• The sample uses selectable text and a regular table. Scanned, merged-cell, or multi-page tables need separate tests.
• This sample has no tax, discounts, or foreign-exchange conversions; do not assume a real invoice uses the same subtotal rules.
• A correct sample does not establish extraction accuracy for other PDFs.