financial-data-entry - SKILL.md Agent Skill

name: financial-data-entry description: > This skill should be used when the user wants to "record financial data", "enter transaction", "add revenue", "input expenses", "record income", "log financial record", "create financial entry", or "add financial figures". It guides the user through structured financial data entry and submission. when_to_use: > Activate when the user needs to manually enter financial data into the system. The skill handles: collecting data from user, validating format and business rules, presenting summary for confirmation, and submitting to the backend database. NOT for Excel bulk uploads — direct the user to the upload feature instead. metadata: author: internal-finance-team version: "2.0" domain: finance compatibility: | Requires: python3 Depends on: POST /api/v1/financial-data endpoint (create single record) allowed-tools: Bash Read Write

Financial Data Entry Skill

Overview

Guide the user through entering a single financial data record into the system. The skill follows a collect → validate → confirm → submit workflow.

The system stores financial data per fiscal year with revenue, expense, and tax breakdowns. Each record belongs to a specific fiscal year and period type (yearly/quarterly/monthly).

Workflow

Step 1: Collect Basic Information

Ask the user for the following required information. If the user provides all info upfront, parse it directly.

Fiscal Year (required):

Ask "Which fiscal year does this data belong to?" if not specified
Must be between 2000 and 2100

Period Type (required):

Options: yearly (default), quarterly, monthly
If period type is quarterly, ask "Which quarter (Q1/Q2/Q3/Q4)?"
If period type is monthly, ask "Which month (1-12)?"
Default to yearly if user doesn't specify

Period Dates (auto-calculated):

For yearly: period_start = {fiscal_year}-01-01, period_end = {fiscal_year}-12-31
For quarterly: period_start = {fiscal_year}-{quarter_start}, period_end = {fiscal_year}-{quarter_end}
For monthly: period_start = {fiscal_year}-{month}-01, period_end = {fiscal_year}-{month}-{last_day}

Step 2: Collect Financial Figures

Ask for the following figures. Only collect what the user can provide — missing fields default to 0.

Revenue (收入):

total_revenue — 营业收入总额 (required if entering revenue data)
taxable_sales — 应税销售额 (optional, subset of revenue)
tax_free_sales — 免税销售额 (optional, subset of revenue)

Expenses (支出):

total_expenses — 总支出 (required if entering expense data)
deductible_expenses — 可扣除费用 (optional)
non_deductible_expenses — 不可扣除费用 (optional)

Tax (税务):

input_tax — 进项税额 (optional)
output_tax — 销项税额 (optional)
vat_rate — 增值税率 (default: 0.13, range: 0-1)
taxable_income — 应纳税所得额 (optional)
corporate_tax_rate — 企业所得税率 (default: 0.25, range: 0-1)

Payroll (工资):

total_payroll — 工资总额 (optional)
special_deductions — 专项扣除 (optional)

Invoice (发票):

total_invoices — 总发票数 (optional)
input_invoice_count — 进项发票数 (optional)
output_invoice_count — 销项发票数 (optional)

Step 3: Validate Data

Run the validation script before submitting:

python scripts/validate_entry.py --params <param_file>

Pass all collected data as the params JSON. The script will:

Validate field types and ranges
Check business rules (revenue >= taxable_sales + tax_free_sales)
Check tax rate ranges (0-1)
Validate date/year consistency

If validation returns errors, show them to the user and ask for corrections.

Step 4: Present Summary and Confirm

Show the user a structured summary:

Financial Data Summary
======================
Fiscal Year: 2025
Period: Yearly (2025-01-01 ~ 2025-12-31)

Revenue:
  Total Revenue:    ¥500,000.00
  Taxable Sales:    ¥450,000.00
  Tax-Free Sales:   ¥50,000.00

Expenses:
  Total Expenses:   ¥300,000.00
  Deductible:       ¥250,000.00
  Non-Deductible:   ¥50,000.00

Tax:
  VAT Rate:         13%
  Output Tax:       ¥58,500.00
  Input Tax:        ¥30,000.00

Calculations:
  Gross Profit:     ¥200,000.00
  Estimated VAT:    ¥28,500.00

Confirm submission? (yes/no)

Ask the user to confirm. If they say no, ask what to change.

Step 5: Submit to Backend

Once confirmed, submit the data:

python scripts/submit_entry.py --params <param_file>

The submit script includes auth context (tenant_id, user_id) passed from the orchestrator. It calls POST /api/v1/financial-data to create the record.

Step 6: Report Result

Report the result back to the user:

On success: Show the record ID and a confirmation message
On failure: Show the error and suggest next steps

Error Handling

Validation errors: Show each error message clearly, ask user to correct
Duplicate record: If a record for the same year/period already exists, inform the user and ask if they want to update it instead
API failure: Retry once, then report the error with details. Do not lose the data — save it and ask user to try again later
Network error: Ask user to check connection and retry

Gotchas

All amounts are in CNY (yuan), no decimals needed for display but accept up to 2 decimal places
Revenue fields must be >= 0 and total_revenue >= taxable_sales + tax_free_sales
Tax rates are expressed as decimals (0.13 = 13%), not percentages
If the user says "I want to enter last year's data", calculate fiscal_year = current_year - 1
For bulk data (many records at once), tell the user to use the Excel upload feature instead
Always ask for confirmation before submitting — financial data is sensitive
Never make up data; if the user doesn't know a value, default it to 0 or skip it
If the user provides "profit" or "net income", calculate it from revenue - expenses, don't accept it as a direct input