Grocery Budget Audit Skill
Food spending is where budgets go to get lied to — partly because it hides
in four places at once (the big shop, the top-ups, the delivery apps, the
work lunches), and partly because every audit turns moral ("we're so bad").
This one doesn't: it's the [[attention-reset]] method pointed at food money
— chosen spending is protected (the Friday takeaway you love stays), and
only the captured spending is the target: the food thrown away, the
convenience markup on autopilot, the brand tax nobody chose, and the
delivery blur that's really a logistics failure wearing a menu.
What This Skill Produces
- A four-stream ledger: big shops / top-ups / delivery-takeaway / food
out, built from whatever evidence exists (receipts, statements, app
order history — the app history is where the surprise lives)
- The leak report: the four leaks quantified for this household, each
with its real cause (waste is usually a planning gap, not gluttony)
- Per-leak fixes with honest expected ranges — labelled as typical
outcomes to verify against next month, never promised
- A keep-the-joy budget: chosen pleasures named and funded, captured
spending targeted, and the one-month re-audit date
Required Inputs
Ask for (if not already provided):
- The evidence: last month's statements/receipts/app histories — or the
honest-guess version, labelled as such, with the audit flagged as
provisional until real numbers arrive
- Household shape: people, work-lunch patterns, who shops, who cooks
- The chosen pleasures to protect — the spending that's working
- What's been tried and how it died (cash envelopes? meal kits?)
Framework
- Assemble the ledger without commentary. Four streams, one month,
totals per stream. The delivery-app order history is mandatory
evidence if apps are used — memory always undercounts it, and the
gap between guess and history is usually the audit's headline.
- Sort chosen from captured. The planned Friday takeaway: chosen —
protected, budgeted, guilt evicted. The 9pm Tuesday order because
nothing was defrosted: captured — a planning gap, addressable. Same
meal, different category; the distinction does all the work.
- Quantify the four leaks. Waste (what got binned — ask for the
honest recurring victims; multiply out) · convenience markup
(top-up-shop pricing, pre-cut everything — the top-up trip count is
the lever, not the lettuce) · brand autopilot (the categories where
own-brand is identical; typically pantry staples, cleaning, basics) ·
the delivery blur (fees + markups + the order-because-tired pattern —
fix is [[meal-prep-os]]'s Thursday plan, not willpower).
- Fix per leak, sized honestly. Each fix names its mechanism and a
typical range ("top-up trips 4→1/week commonly saves 10-20% of that
stream — verify against your month"). No global "slash your bill 50%"
claims; the re-audit is the referee.
- Set the budget with the joy line in it. Streams budgeted with the
chosen items explicitly inside; one flexible buffer; the rule that
protects the system: overspend gets investigated (which leak?) not
punished. Re-audit in one month against real numbers.
Output Format
## The ledger — [month]
| Stream | Total | Biggest line | Chosen / captured split |
[The guess-vs-history gap, if apps were involved]
## Leak report
| Leak | This household's number | Real cause |
## Fixes (ranges to verify, not promises)
[Per leak: mechanism · typical range · this household's version]
## The budget (joy included)
[Per stream · the protected list · buffer · the investigate-don't-punish
rule · re-audit date]
Quality Checks
Anti-Patterns
Related
[[meal-prep-os]] is the fix for half the leaks; [[attention-reset]] — the
same chosen/captured method, different currency; [[debt-payoff]] when the
found money needs a destination.
1---2name: grocery-budget-audit3description: Find where the food money actually goes — a no-shame ledger built from real receipts/statements, the four leak categories (waste, convenience markup, brand autopilot, the takeaway blur), a per-leak fix with realistic savings ranges, and a target budget that survives real life. Use when someone says 'we spend how much on food?!', 'audit my grocery spending', 'cut our food bill', or takeaway guilt is the household argument. Produces the ledger, the leak report, and a keep-the-joy budget.4---5
6# Grocery Budget Audit Skill
7
8Food spending is where budgets go to get lied to — partly because it hides
9in four places at once (the big shop, the top-ups, the delivery apps, the
10work lunches), and partly because every audit turns moral ("we're so bad").
11This one doesn't: it's the [[attention-reset]] method pointed at food money
12— chosen spending is protected (the Friday takeaway you love stays), and
13only the *captured* spending is the target: the food thrown away, the
14convenience markup on autopilot, the brand tax nobody chose, and the
15delivery blur that's really a logistics failure wearing a menu.
16
17## What This Skill Produces
18
19- A **four-stream ledger**: big shops / top-ups / delivery-takeaway / food
20 out, built from whatever evidence exists (receipts, statements, app
21 order history — the app history is where the surprise lives)
22- The **leak report**: the four leaks quantified for this household, each
23 with its real cause (waste is usually a planning gap, not gluttony)
24- **Per-leak fixes** with honest expected ranges — labelled as typical
25 outcomes to verify against next month, never promised
26- A **keep-the-joy budget**: chosen pleasures named and funded, captured
27 spending targeted, and the one-month re-audit date
28
29## Required Inputs
30
31Ask for (if not already provided):
32- The evidence: last month's statements/receipts/app histories — or the
33 honest-guess version, labelled as such, with the audit flagged as
34 provisional until real numbers arrive
35- Household shape: people, work-lunch patterns, who shops, who cooks
36- The chosen pleasures to protect — the spending that's working
37- What's been tried and how it died (cash envelopes? meal kits?)
38
39## Framework
40
411. **Assemble the ledger without commentary.** Four streams, one month,
42 totals per stream. The delivery-app order history is mandatory
43 evidence if apps are used — memory *always* undercounts it, and the
44 gap between guess and history is usually the audit's headline.
452. **Sort chosen from captured.** The planned Friday takeaway: chosen —
46 protected, budgeted, guilt evicted. The 9pm Tuesday order because
47 nothing was defrosted: captured — a planning gap, addressable. Same
48 meal, different category; the distinction does all the work.
493. **Quantify the four leaks.** Waste (what got binned — ask for the
50 honest recurring victims; multiply out) · convenience markup
51 (top-up-shop pricing, pre-cut everything — the top-up *trip count* is
52 the lever, not the lettuce) · brand autopilot (the categories where
53 own-brand is identical; typically pantry staples, cleaning, basics) ·
54 the delivery blur (fees + markups + the order-because-tired pattern —
55 fix is [[meal-prep-os]]'s Thursday plan, not willpower).
564. **Fix per leak, sized honestly.** Each fix names its mechanism and a
57 typical range ("top-up trips 4→1/week commonly saves 10-20% of that
58 stream — verify against your month"). No global "slash your bill 50%"
59 claims; the re-audit is the referee.
605. **Set the budget with the joy line in it.** Streams budgeted with the
61 chosen items explicitly inside; one flexible buffer; the rule that
62 protects the system: overspend gets *investigated* (which leak?) not
63 punished. Re-audit in one month against real numbers.
64
65## Output Format
66
67```
68## The ledger — [month]
69| Stream | Total | Biggest line | Chosen / captured split |
70[The guess-vs-history gap, if apps were involved]
71
72## Leak report
73| Leak | This household's number | Real cause |
74
75## Fixes (ranges to verify, not promises)
76[Per leak: mechanism · typical range · this household's version]
77
78## The budget (joy included)
79[Per stream · the protected list · buffer · the investigate-don't-punish
80rule · re-audit date]
81```
82
83## Quality Checks
84
85- [ ] All four streams present; delivery history requested by name when
86 apps are in play
87- [ ] Chosen spending is explicitly protected and appears IN the budget
88- [ ] Every savings number is a labelled range-to-verify, never a promise
89- [ ] Waste analysis names actual recurring victims from the inputs, not
90 hypothetical lettuce
91- [ ] The tone check: zero shame anywhere — data, causes, mechanisms,
92 re-audit
93
94## Anti-Patterns
95
96- [ ] Do not moralize any line of the ledger — the audit that shames is
97 the audit that gets abandoned
98- [ ] Do not target the chosen pleasures — cutting the loved takeaway is
99 how the whole budget loses its constituency
100- [ ] Do not recommend extreme-couponing/13-store optimization — time is
101 a cost; the fixes here are structural, not heroic
102- [ ] Do not assert current prices or "average household" comparisons as
103 fact — this household's month is the only benchmark
104- [ ] Do not skip the re-audit date; unverified fixes become folklore
105
106## Related
107
108[[meal-prep-os]] is the fix for half the leaks; [[attention-reset]] — the
109same chosen/captured method, different currency; [[debt-payoff]] when the
110found money needs a destination.
Mohit Aggarwal