reading-withholding

Withholding Tax Slip Image Reading

A skill that reads an image of a withholding tax slip and returns it as structured data.

For PDF files

The file is PDF (.pdf), attempt text extraction before image OCR.

1. shinkoku pdf extract-text --file-path <path> Run
2. If the extracted text contains the necessary information (payment amount, withholding tax amount, etc.), generate structured data from the text
3. If the text is insufficient (scanned PDF, etc.), use shinkoku pdf to-image --file-path <path> --output-dir <dir> to convert to PNG and proceed to the following image reading flow

Image Reading Method

Recommended: Dual verification (two parallel contexts)

To improve accuracy, read the same image in two independent contexts in parallel and cross-check the results.

1. Perform two independent reads: In environments where subagents are available, launch two subagents in parallel, each reading the image independently. Pass the following "Basic Rules" and "Output Format" to each subagent as a prompt, and specify the image file path.

2. Result cross-check: Compare the major fields (amounts, etc.) from both reading results.

3. If they match: Adopt as is. Report that "the results matched across two independent reads."

4. If they do not match: Present the original image path and both results to the user, and have them choose the correct one:

   • Explicitly indicate the fields that differ
   • Present three options: Adopt A / Adopt B / Manual input

Fallback (when subagents are not supported)

In environments where subagents are unavailable, read using the following procedure:

1. Read the image file directly with the Read tool
2. Extract data according to the following "Basic Rules" and "Output Format"
3. Present the extraction results to the user andalways request visual confirmation

⚠ Since dual verification is unavailable, always request visual confirmation from the user.

Basic Rules

• Read image files with the Read tool (Claude Vision automatically recognizes images)
• Always return amounts as int (integer in yen). Remove commas and "円"
• Return dates in YYYY-MM-DD format
• Convert Japanese era names to Gregorian years (Reiwa 7 → 2025, Reiwa 6 → 2024, Heisei 31 → 2019)
• For fields that cannot be read, use UNKNOWN (string) or 0 (amount)
• If multiple files are provided, process them all in order and return them together

Output Format

Read the image and return it in the following format:

---WITHHOLDING_DATA---
payer_name: 支払者名
payment_amount: 支払金額（int）
withheld_tax: 源泉徴収税額（int）
social_insurance: 社会保険料等の金額（int）
life_insurance_deduction: 生命保険料の控除額（int）
earthquake_insurance_deduction: 地震保険料の控除額（int）
housing_loan_deduction: 住宅借入金等特別控除の額（int）
life_insurance_detail:
  general_new: 一般の新保険料（int）
  general_old: 一般の旧保険料（int）
  medical_care: 介護医療保険料（int）
  annuity_new: 個人年金の新保険料（int）
  annuity_old: 個人年金の旧保険料（int）
---END---

Extraction points

• Extract the "Payment Amount" field (total salary income) with the highest priority
• Accurately read the "Withholding Tax Amount" field
• Read the "Social Insurance Premiums, etc. Amount" field
• For life insurance premium deductions, check the breakdown of the old and new systems and the 3 categories (general/nursing care medical/individual annuity)
• Extract earthquake insurance premium deductions and housing loan deductions only if they are stated
• Extract the payer's name (company name)
• Items that are not stated shall be 0

Processing multiple files

When multiple file paths are specified:

1. Obtain the file list with the Glob tool (when a pattern is specified)
2. Read each file in order with the Read tool
3. Return the results of all files together (state the file name before each result)

## file1.jpg
---WITHHOLDING_DATA---
...
---END---

## file2.jpg
---WITHHOLDING_DATA---
...
---END---