monthly-close-checklist-and-reconciliation-prep

name	monthly-close-checklist-and-reconciliation-prep
description	Turns a pasted bank export or transaction list into a monthly-close checklist, categorization draft with disputed flags, and an accountant/investor summary — no SaaS needed. Triggers: 'monthly close checklist', 'reconcile transactions', 'чек-лист месячного закрытия', 'сверь транзакции'.
version	1.0.0

Monthly Close Checklist and Reconciliation Prep

This skill turns a pasted transaction list or bank export (CSV rows or markdown table) plus a brief business context into three ready-to-use artifacts: a monthly-close checklist, a categorization draft with disputed flags, and an accountant/investor-ready summary — no accounting SaaS or integrations required.

Input:

• Transaction list: CSV rows or markdown table (minimum columns: date, counterparty/description, amount)
• Business context: entity type (ИП / ООО / sole-prop), primary currency, brief description of revenue model

Output:

• Artifact 1 — Monthly-Close Checklist (locale-adapted)
• Artifact 2 — Categorization Draft with Disputed Flags
• Artifact 3 — Accountant Snapshot + Investor Narrative

---

Language Detection

Detect the user's language from their message:

• If Russian (or contains Cyrillic): respond in Russian
• If English (or other Latin-script language): respond in English
• If ambiguous: respond in the language of the trigger phrase used

---

Instructions

Step 1: Validate Input

1. Check that a transaction list is present in the user's message (CSV rows, markdown table, or plain list with date + counterparty + amount)

   • If no transaction list detected: stop. Return bilingual prompt:

     "Для работы нужен список транзакций — вставьте CSV-строки или таблицу в формате Markdown. / I need a transaction list — paste CSV rows or a Markdown table."

2. Check that at least one parseable data row exists (not just headers or empty rows)

   • If zero data rows found: stop. Return in user's language:

     "Список не содержит транзакций. Убедитесь, что данные включают хотя бы одну строку с датой, контрагентом и суммой. / The list contains no transactions. Include at least one row with date, counterparty, and amount."

3. Note any missing business context (entity type, currency, revenue model)

   • If absent: proceed with generic international defaults; note assumptions in the output preamble

Step 2: Map Transaction Columns

1. Identify the date column (date-like values or headers: "date", "дата", "datum")
2. Identify the counterparty/description column (free text, merchant names, or descriptions)
3. Identify the amount column (numeric; note whether debit/credit are separate or a single signed column)
4. Identify the direction column (income/expense) if present; otherwise infer from amount sign

Edge Case — Non-standard column names:

• Attempt heuristic detection; list the detected column mapping in the output preamble
• Then ask the user to confirm the mapping before proceeding: "Пожалуйста, подтвердите, что колонки определены верно / Please confirm the detected column mapping is correct: [mapping table]. Reply 'yes' to continue or correct any column names."
• Do not generate the three artifacts until the user confirms or corrects the mapping

Edge Case — Mixed currencies:

• Process each currency separately
• Flag multi-currency rows in the Disputed section; add note: "FX conversion is out of scope"

Step 3: Generate Categorization Draft and Flag Disputed Transactions

1. Scan counterparty names and descriptions to group transactions by pattern (recurring SaaS subscriptions, salary payouts, tax transfers, client payments)
2. Propose a draft category label for each group:
   • Income: Revenue/Consulting, Revenue/Product, Other Income
   • Expense: OPEX/Software, OPEX/Payroll, OPEX/Office, Tax, Transfer, Other
3. Mark a transaction as ⚠ DISPUTED when:
   • Counterparty is ambiguous or matches no clear pattern
   • Amount is unusually large relative to others in the same group
   • Transaction direction cannot be determined from the data
4. Build the categorization table (see Output Format)

Step 4: Generate Monthly-Close Checklist

1. Start with standard items for all entity types:

   • Verify bank balance matches internal records
   • Confirm all transactions are categorized (zero uncategorized rows)
   • Review and approve all ⚠ DISPUTED items
   • Check outstanding receivables (unpaid invoices this month)
   • Check outstanding payables (bills due)
   • Confirm payroll is reflected in records
   • Verify tax payments and withholdings are current

2. Add locale-specific items based on entity type:

   • ИП (Russia/CIS, УСН):
     • Обновить КУДиР записями за месяц (колонка «Доходы» для УСН 6%)
     • Сверить остатки по банку с данными из личного кабинета
     • Проверить акт сверки с ключевыми контрагентами при наличии дебиторки
     • Уточнить срок авансового платежа (Q1: 28 апр, Q2: 28 июл, Q3: 28 окт)
   • ООО (Russia/CIS):
     • Проверить подотчётные суммы (авансовые отчёты сотрудников)
     • Подтвердить корректность записей по НДС (для плательщиков НДС)
   • Sole-prop / EN default:
     • Review mileage log or home-office deduction documentation
     • Confirm quarterly estimated tax payment is scheduled

3. Append 1–3 custom items derived from the most critical ⚠ DISPUTED rows in Step 3 (e.g., "Clarify [date] [counterparty] [amount] before handoff")

Step 5: Generate Accountant Snapshot

1. Compute totals from the categorized table:
   • Total income by category; total expenses by category; net cash flow (income − expenses)
2. List all ⚠ DISPUTED items (date, counterparty, amount, reason for flag)
3. List open questions for ambiguous categorization (formulate as clear yes/no or clarification questions)

Step 6: Generate Investor Narrative

1. Write a 3–5 sentence plain-language narrative covering:
   • Revenue for the month (and MoM trend if prior context was provided)
   • Cash position after expenses
   • 1–2 key risks or open items (e.g., large disputed transaction, delayed receivable)
2. Keep language accessible to a non-accountant reader — no formulas, no accounting jargon
3. One paragraph only

---

Negative Cases

• No transaction list provided: Stop. Return bilingual prompt (Step 1).
• Zero parseable rows: Stop. Return bilingual message (Step 1).
• User requests SaaS integration (1С, QuickBooks, Xero, etc.): Decline. State: "Этот скил работает только с вставленными данными и не подключается к внешним системам. / This skill works only with pasted data and does not connect to external systems."
• List >200 rows: Process all rows for categorization; produce aggregate totals in investor narrative; note total row count in preamble.

---

Output Format

Single markdown response with three artifact sections plus an optional preamble:

> **Assumptions / Допущения:** [column mapping notes, missing-context defaults, currency split if applicable]

---

## Чек-лист месячного закрытия / Monthly-Close Checklist
**Период / Period:** [Month YYYY]
**Тип субъекта / Entity type:** [ИП / ООО / Sole-prop / Unknown]

- [ ] Item 1
- [ ] Item 2

---

## Правила категоризации / Categorization Draft
| Дата / Date | Контрагент / Counterparty | Сумма / Amount | Категория / Category | Статус / Status |
|---|---|---|---|---|
| YYYY-MM-DD | Name | +/- amount [CCY] | Category/Subcategory | ✅ / ⚠ DISPUTED |

> Спорных транзакций / Disputed: [N]. Отметьте их перед передачей бухгалтеру / Review before handoff.

---

## Резюме для бухгалтера / Accountant Snapshot
**Период / Period:** [Month YYYY]
**Доходы / Income:** [total] | **Расходы / Expenses:** [total] | **Нетто / Net:** [net]

**По категориям / By category:**
| Категория / Category | Сумма / Amount |
|---|---|
| Revenue/Consulting | + |

**Спорные позиции / Disputed items:**
- YYYY-MM-DD | [Counterparty] | [Amount] — [reason]

**Открытые вопросы / Open questions:**
1. [Question for accountant]

---

## Резюме для инвестора / Investor Narrative
[3–5 sentence plain-language paragraph: revenue, cash position, 1–2 key risks]

Formatting rules:

• All four section headers are bilingual (RU / EN) regardless of detected language
• Body text (list items, table cells, narrative) is in the detected user language
• ⚠ DISPUTED tag is always in English as a universal visual signal
• Net cash flow = total income − total expenses derived from the categorized table