| name | family-budget-il |
| description | Set up and run a shared family budget planner as a Google-Sheets-ready workbook, with natural-language expense logging through chat. Defaults to Israel (Hebrew/RTL, shekels, Israeli categories, HMOs, 15th-to-15th salary cycle, provident/pension funds, benefit clubs) but works for any country. Use whenever the user wants to build a family/household budget, create a shared budget spreadsheet, plan monthly spending by category, log expenses/salary/ bills in plain language ("spent 250 at Shufersal", "arnona 1200", "salary came in"), see where the family stands vs. budget, track net worth and investments alongside spending, or get discount/benefit-club recommendations. Trigger on "ืชืงืฆืื ืืฉืคืืชื", "ืืชืื ื ืชืงืฆืื", "ื ืืืื ืืืฆืืืช ืืฉืคืื", "family budget", "household budget", "budget planner", "track our spending", "shared budget sheet", even without the word "skill". Prefer this over a plain spreadsheet whenever the budget is for a household and ongoing logging is implied. |
| license | Proprietary |
Family Budget Planner (Israel-first)
Build a shared, Google-Sheets-ready budget workbook for a household, then let
the family keep it current by typing expenses in plain Hebrew or English โ
no bank connections, no app to install, no data uploads required.
The design goal is lightweight: one workbook the family shares via Google
Sheets, plus chat-based logging. Do not turn this into a system that forces
the family to connect every bank, card, and broker.
How it works (the two-phase model)
- Setup (once) โ interview the family, then generate one
.xlsx workbook.
They upload it to Google Drive โ Open with Google Sheets โ Share with the
family. That sheet is now their live, shared budget.
- Logging (ongoing, through chat) โ the family tells Claude what happened
("ืฉืืืืชื 250 ืืฉืืคืจืกื", "ื ืื ืกื ืืฉืืืจืช 18,000", "ืงื ืก ืื ืื 100"). Claude returns
clean, categorized rows ready to paste into the ืชื ืืขืืช / Transactions tab.
The dashboard recalculates itself. If the Google Drive/Sheets connector is
available, append the rows directly instead of asking them to paste.
Phase 1 โ Setup interview
Run the wizard the first time, or whenever the user says "set up", "new budget",
"start a family budget". Ask in the user's language (Hebrew if they wrote in
Hebrew). Keep it to a few grouped questions โ do not interrogate one field at a
time. Use the ask_user_input tool for the multiple-choice ones (country,
budget cycle, currency) so it is one tap on mobile.
Collect:
- Family members โ names and (optionally) who earns / who has a card.
- Country & currency โ default Israel / โช (ILS). If another country, adjust
currency symbol, swap Israeli-specific categories (HMOs, ืืจื ืื ื, ืืืืื ืืืืื)
for local equivalents, and skip the Israeli clubs section.
- Income sources โ salary(ies), freelance, rental, allowances, benefits.
Capture the salary arrival day (e.g. the 10th) โ it drives the cycle.
- Budget cycle โ Israel default is 15th-to-15th (salary-aligned), since
Israeli salaries land a few days before month-end and a 1st-to-1st cycle
splits the salary from its expenses. Offer: salary-day cycle / 15th-to-15th /
calendar month (1st-to-1st).
- Fixed monthly expenses โ rent/mortgage, ืืจื ืื ื, ืืขื ืืืช, utilities,
internet, phone, insurances, subscriptions, loan repayments, tuition/ืืืืื.
- Accounts โ bank accounts and credit cards (just labels + last 4 digits;
no numbers, no credentials).
- Monthly budget targets โ per category group. If they don't know, seed
sensible defaults from their fixed expenses and ask them to adjust later.
- Savings & investments โ ืงืจื ืืฉืชืืืืช, ืงืจื ืคื ืกืื, ืงืืคืช ืืื, ืื ืืืช/ETF,
savings โ for the net-worth tab. These are asset transfers, not spending
(see budget-logic).
- Goals โ what is the family saving for? (trip, emergency fund, apartment,
debt payoffโฆ). For each: a name, target amount, target date, and how much is
already saved. These seed the ืืขืืื / Goals tab, which computes the
required โช/month and flags when goals together exceed the monthly surplus.
- Benefit / discount clubs (Israel) โ ask which they belong to (ืืืจ, ืืืขืืื
ืืกืชืืจืืช / ืืขื ืขืืืืื, ืืืขืืื ื ืืฉืจืื, ืชื ืืืก/Cibus/Pluxee, supermarket clubs,
etc.). Used later for recommendations. See
references/israeli-clubs.md.
Write the answers to a config JSON, then build the workbook:
python scripts/build_workbook.py --config /home/claude/budget_config.json \
--out /mnt/user-data/outputs/family_budget.xlsx
python scripts/recalc.py /mnt/user-data/outputs/family_budget.xlsx
The config schema and a worked example are in assets/config.example.json.
Read references/categories.md for the full category tree the workbook seeds,
and references/budget-logic.md for how the dashboard formulas must behave.
After building, present the file and give the family three short steps:
upload to Google Drive โ Open with Google Sheets โ Share with the family.
Phase 2 โ Natural-language logging
This is the everyday path. The family sends messages like:
- "ืฉืืืืชื 250 ืืฉืืคืจืกื ืืขืื 40 ืืืจืืื"
- "arnona 1,247 came out today"
- "salary 18,500 net, and Maya's freelance 3,200"
- "parking fine 100, and 600 to the dentist"
- "ืืขืืจืชื 2,000 ืืงืจื ืืฉืชืืืืช"
For each message:
- Parse every line item โ amount, currency, merchant/description, and date
(default to today if unstated). Follow
references/nl-parsing.md.
- Classify each into the taxonomy: pick type (income / expense / transfer
/ investment) and category group + category. Use merchantโcategory mapping
from
references/nl-parsing.md; when unsure, mark category as
ืื ืืกืืื / Uncategorized rather than guessing wildly.
- Set the budget flag โ
exclude_from_budget = TRUE for investments,
savings, and internal transfers (asset moves, not consumption). See
budget-logic.
- Avoid double-counting credit cards โ if they log both a card's monthly
bank deduction and its line items, keep the line items and mark the bank
deduction as a reconciliation/transfer. See budget-logic.
- Output rows in the exact Transactions column order, as a clean block the
user can paste. Echo a one-line confirmation of what was logged and the
running cycle balance if known. Then stop โ don't over-explain.
Transactions column order (must match the workbook):
Date | Member | Account | Type | Group | Category | Description | Amount | ExcludeFromBudget | Notes
If the Google Drive/Sheets connector is connected, offer to append the rows to
their sheet directly instead of pasting. Confirm the sheet name first.
Phase 3 โ Goals, "where do we stand?", and recommendations
When the user asks how they're doing ("ืืืคื ืื ืื ื ืขืืืืื?", "how's the budget?",
"can we afford X?"):
- Summarize the current cycle: total income, operating spend, savings rate,
and the 3โ4 categories most over/under budget. Lead with the number, keep it
to a screenful. Offer a chart only if they want detail.
- Distinguish ืชืงืฆืื ืฉืืืฃ (operating budget) from ืชืืจืื ืืืื (total cash
flow) โ default to operating. See budget-logic.
Goals live on the ืืขืืื / Goals tab. When the user sets or asks about a
goal ("we want โช30k for a trip by July", "are we on track for the down payment?"):
- Make sure the goal row exists (name, target, target date, saved). The tab
auto-computes months left, still-needed, and required โช/month.
- Read the Gap cell (monthly surplus โ total required/month). If negative,
say plainly that the goals together outpace the surplus and offer the levers:
push a date out, trim a category, or raise income. Tie it to the savings rate.
Recommendations live on the ืืืืฆืืช / Recommendations tab. The top of that
tab auto-computes insights (biggest category, most over budget, subscriptions
total, savings rate). The action board below is yours to fill:
- Read
references/israeli-clubs.md. Map the family's declared clubs and their
biggest categories to concrete actions, and write rows into the action
board: Area, Action, Est. โช/month saving, Status. Web-search the specific
clubs they named for current benefits before quoting any number โ perks
change often, so never state a current discount from memory.
- The tab totals the potential saving and shows the improved savings rate if
they action the "ืืืืฆืืข" items โ use that as the motivating headline.
- Be concrete and non-pushy. Note you're not a financial advisor for anything
involving investing or pensions; hand pension/fund questions to the
gemelnet-advisor skill if available.
Reference files
Read these as needed โ don't load them all up front.
references/categories.md โ the full bilingual Israeli category taxonomy (18
groups, ~70 categories) that the workbook seeds. Read before building or when
categorizing an unusual item.
references/budget-logic.md โ the calculation rules reused from the user's
FinanceOS project: operating-vs-total cash flow, the ExcludeFromBudget flag,
credit-card reconciliation, salary-aligned cycles, savings rate, net worth.
Read before building the dashboard or when a logging edge case is unclear.
references/nl-parsing.md โ how to turn Hebrew/English plain-language messages
into transaction rows, including Israeli merchantโcategory mappings and
currency/amount parsing. Read whenever logging transactions.
references/israeli-clubs.md โ major Israeli benefit/discount programs and how
to turn club membership into spending recommendations. Read for Phase 3.
Scripts
scripts/build_workbook.py โ generates the Google-Sheets-ready .xlsx from a
config JSON. All dashboard cells are live formulas (SUMIFS over Transactions),
so the sheet stays correct as rows are added. Dropdowns use data validation so
they survive the import to Google Sheets.
scripts/recalc.py โ (copied from the xlsx skill) recalculates formulas and
reports any formula errors. Run after building; fix until zero errors.
Guardrails
- Never ask for or store account numbers, card numbers, passwords, or
credentials. Labels and last-4 only.
- Keep every workbook formula error-free (no #REF!, #DIV/0!, #VALUE!).
- Don't fabricate current club discounts or fund yields โ search first.
- For anything touching pensions/provident funds or investing decisions, hand off
to the
gemelnet-advisor skill if available, and note you're not a financial
advisor.