aisle/tools/Buy at the right price - agent workflow | AISLE
free to read Buy 10 runs →

Buy it at this price · workflow

Buy at the right price

Check the real price history, the coupon stack and the landed cost before buying.

agent Buying agenttime 15-30 minsteps 17

Inputs

ask once, then run

keytypewhat it is
product requiredstring: name and exact variantProduct name or retailer URL
targetnumber: price you would payTarget price
e.g. 59
zipstring: delivery postcodeDelivery ZIP or country
e.g. US

Run it

17 steps

1

Normalise the product

resolve_input({ input: "{{product}}" }) resolve_input

Turn the product or URL into one canonical product and variant. Size, colour and pack count change the price.

returnsproduct, variant, category, retailer

2

Check the market context

shelf_search({ q: "{{product}}", sort: "views", limit: 10 }) shelf_search

See whether the product is currently being promoted organically. Promotion often predicts a discount window.

returnsrecent rows, formats, view velocity

3

Pull price history

KEEPA / CAMELCAMELCAMEL / CHANGEDETECTION

Read the 30, 90 and 365-day low and the last price change. Use the paid history source only if the free one lacks the marketplace.

returnscurrent, 30d low, 90d low, last change date

4

Check the coupon stack

RETAILER CHECKOUT

Test the visible coupon, the newsletter code and the cart-level discount. Keep only codes that apply at checkout; a code that fails is not a saving.

returnsworking codes, order of application, final cart price

5

Add landed cost

LANDED COST

Add shipping, tax, duty and returns risk. For a cross-border retailer, show the duty assumption separately.

returnslanded cost, per-unit cost at pack size

6

Decide

BUY OR WAIT

Compare the landed cost to the target and the 90-day low. If today is within 5% of the low, buy; otherwise set an alert and say when to re-check.

returnsbuy/wait, trigger price, re-check date

7

Timestamp the price

MANUAL

Record the current price with the time and the page it came from. A price without a timestamp is a rumour.

returnsprice, currency, timestamp, URL

8

Pull the price history

MANUAL

Read the price history for the same SKU and size. Note the 90-day low, the 90-day median and the last time it hit the low.

returnslow, median, last-low date

9

Stack the coupons

MANUAL

Test each coupon against the cart, one at a time, and record which ones actually apply. A coupon that shows in the banner but fails at checkout is not a coupon.

returnsapplied coupon, discount, final cart price

10

Add the landed cost

MANUAL CALC

Add shipping, tax and any membership cost to the cart price. Compare that number to the target, not the sticker price.

returnslanded cost, delta to target

11

Decide and write the trigger

MANUAL

State buy or wait, the price that would change the answer, and the date the check expires. A wait without a trigger is procrastination.

returnsdecision, trigger price, expiry date

12

Check the seller

MANUAL

Record who ships it, the return window, the restocking fee and the warranty. A cheap price from a seller with a seven-day window and a 20 percent restocking fee is not cheap.

returnsseller terms: shipper, window, fee, warranty

13

Compare the substitute

MANUAL

Name the nearest alternative that does the same job at a different price and put it on the same landed-cost line. A decision with one option is not a decision.

returnstwo options on one landed-cost comparison

14

Set the price watch

MANUAL

Write the threshold and the expiry: the price at which you buy, and the date the watch ends. An alert with no expiry becomes a subscription to wanting.

returnsthreshold, expiry, where the alert lives

15

Photograph the deal

MANUAL

Screenshot the cart with the date, the time and the applied coupon visible. If the price moves at checkout, this image is the only proof of what you were shown.

returnscart image with date, time and coupon visible

16

Read the returns policy as a cost

MANUAL

Turn the return window, the restocking fee and the return shipping into a percentage of the price. A generous price with a punitive return is a worse deal than it looks on the sticker.

returnsreturn cost as a percentage of the price

17

Set the replacement cycle

MANUAL

State how long the thing should last and divide the landed cost by the months. The per-month number is what compares to the substitute, not the sticker.

returnsexpected life, cost per month, comparison

How this run is judged

the judgement calls

A price is a claim with a timestamp, and the job is to keep the two together. The banner is marketing, the cart is evidence, and the landed cost is the only number that compares to your target. Read the seller as carefully as the price: who ships, what the return window really is, what the restocking fee takes back. The judgement call is the substitute: if the nearest alternative at a different price does the same job, the decision is between two products, not between a price and a target. Price the thing per month of use, because a cheap object replaced twice a year is the expensive one.

Files that land on disk

2 files

fileschema
price-check.mdsku, price, timestamp, history, coupons, landed cost, decision
triggers.jsontarget price, trigger price, expiry, source

Cost and the stop

before the spend

15-30 minutes and no tooling. The real cost is the temptation to buy at the sticker price; the saving is the delta between the sticker and the landed cost at the recent low.

Stop condition. Stop when the SKU or the size cannot be confirmed: a price for a different variant is worse than no price.

What makes this run fail

3 failure modes

failurecause in the recordfix
The price was a different variantThe size or SKU was never confirmedConfirm the variant on the cart, not on the search result
The coupon did not applyThe banner was read instead of the cartTest every code in the cart and record the ones that failed
The wait never endsNo trigger price and no expiryWrite both before the tab closes

Next

the output this run feeds

Validate demand — What you just paid is the first demand signal; the next question is whether anyone else will.

The prompt

paste into the agent

You are the Buying agent running the AISLE workflow "Buy at the right price".

MISSION
Decide buy or wait using price history, not a countdown timer. Give the current price, the recent low, the coupon stack that actually applies, and the landed cost. Never state a live price without a timestamp, and never call a banner a discount until the cart agrees.

HOW THIS RUN IS JUDGED
A price is a claim with a timestamp, and the job is to keep the two together. The banner is marketing, the cart is evidence, and the landed cost is the only number that compares to your target. Read the seller as carefully as the price: who ships, what the return window really is, what the restocking fee takes back. The judgement call is the substitute: if the nearest alternative at a different price does the same job, the decision is between two products, not between a price and a target. Price the thing per month of use, because a cheap object replaced twice a year is the expensive one.

INPUTS
- product (required): Product name or retailer URL [string: name and exact variant]
- target (optional): Target price [number: price you would pay]. Example: 59
- zip (optional): Delivery ZIP or country [string: delivery postcode]. Example: US
Ask once for any missing required input. Do not guess a URL, a product or a price.

LIVE STACK FOR THIS JOB (from the AISLE index; join links are the verified vendor pages)
Call stack_get({ id: "buy" }) for the full record with every option for each seat.
- Retailer: Amazon Associates | $0: Google Shopping | also does this job: PriceHistory, eBay Partner Network, Walmart Creator, Target Partners
- Price history: $0: CamelCamelCamel | also does this job: PriceHistory
- Buying to resell: Zendrop | $0: Alibaba | also does this job: CJdropshipping, Faire, 1688, Made-in-China

TOOLS, IN ORDER
1. Normalise the product
   Call: resolve_input({ input: "{{product}}" })
   Do: Turn the product or URL into one canonical product and variant. Size, colour and pack count change the price.
   Return: product, variant, category, retailer
2. Check the market context
   Call: shelf_search({ q: "{{product}}", sort: "views", limit: 10 })
   Do: See whether the product is currently being promoted organically. Promotion often predicts a discount window.
   Return: recent rows, formats, view velocity
3. Pull price history
   Call: KEEPA / CAMELCAMELCAMEL / CHANGEDETECTION
   Do: Read the 30, 90 and 365-day low and the last price change. Use the paid history source only if the free one lacks the marketplace.
   Return: current, 30d low, 90d low, last change date
4. Check the coupon stack
   Call: RETAILER CHECKOUT
   Do: Test the visible coupon, the newsletter code and the cart-level discount. Keep only codes that apply at checkout; a code that fails is not a saving.
   Return: working codes, order of application, final cart price
5. Add landed cost
   Call: LANDED COST
   Do: Add shipping, tax, duty and returns risk. For a cross-border retailer, show the duty assumption separately.
   Return: landed cost, per-unit cost at pack size
6. Decide
   Call: BUY OR WAIT
   Do: Compare the landed cost to the target and the 90-day low. If today is within 5% of the low, buy; otherwise set an alert and say when to re-check.
   Return: buy/wait, trigger price, re-check date
7. Timestamp the price
   Call: MANUAL
   Do: Record the current price with the time and the page it came from. A price without a timestamp is a rumour.
   Return: price, currency, timestamp, URL
8. Pull the price history
   Call: MANUAL
   Do: Read the price history for the same SKU and size. Note the 90-day low, the 90-day median and the last time it hit the low.
   Return: low, median, last-low date
9. Stack the coupons
   Call: MANUAL
   Do: Test each coupon against the cart, one at a time, and record which ones actually apply. A coupon that shows in the banner but fails at checkout is not a coupon.
   Return: applied coupon, discount, final cart price
10. Add the landed cost
   Call: MANUAL CALC
   Do: Add shipping, tax and any membership cost to the cart price. Compare that number to the target, not the sticker price.
   Return: landed cost, delta to target
11. Decide and write the trigger
   Call: MANUAL
   Do: State buy or wait, the price that would change the answer, and the date the check expires. A wait without a trigger is procrastination.
   Return: decision, trigger price, expiry date
12. Check the seller
   Call: MANUAL
   Do: Record who ships it, the return window, the restocking fee and the warranty. A cheap price from a seller with a seven-day window and a 20 percent restocking fee is not cheap.
   Return: seller terms: shipper, window, fee, warranty
13. Compare the substitute
   Call: MANUAL
   Do: Name the nearest alternative that does the same job at a different price and put it on the same landed-cost line. A decision with one option is not a decision.
   Return: two options on one landed-cost comparison
14. Set the price watch
   Call: MANUAL
   Do: Write the threshold and the expiry: the price at which you buy, and the date the watch ends. An alert with no expiry becomes a subscription to wanting.
   Return: threshold, expiry, where the alert lives
15. Photograph the deal
   Call: MANUAL
   Do: Screenshot the cart with the date, the time and the applied coupon visible. If the price moves at checkout, this image is the only proof of what you were shown.
   Return: cart image with date, time and coupon visible
16. Read the returns policy as a cost
   Call: MANUAL
   Do: Turn the return window, the restocking fee and the return shipping into a percentage of the price. A generous price with a punitive return is a worse deal than it looks on the sticker.
   Return: return cost as a percentage of the price
17. Set the replacement cycle
   Call: MANUAL
   Do: State how long the thing should last and divide the landed cost by the months. The per-month number is what compares to the substitute, not the sticker.
   Return: expected life, cost per month, comparison

FILES THAT LAND ON DISK
- price-check.md - sku, price, timestamp, history, coupons, landed cost, decision
- triggers.json - target price, trigger price, expiry, source
Write every file with the name above. A deliverable that exists only in chat is not delivered.

COST AND TIME
15-30 minutes and no tooling. The real cost is the temptation to buy at the sticker price; the saving is the delta between the sticker and the landed cost at the recent low.

STOP CONDITION
Stop when the SKU or the size cannot be confirmed: a price for a different variant is worse than no price.

RESEARCH HARNESS
You are an operator, not a form. Run this loop before you act:
1. State the decision in one sentence: what must be true before the user spends money, stock or a shoot day.
2. Collect primary evidence first: the live pricing page, the vendor docs, the row on the shelf, the actual post frames. Record the URL and the date. A listicle is a lead, not proof.
3. Name the assumption that would break the answer. Test that one first.
4. Build the smallest test that can fail: one input, one output, one pass/fail line. Do not scale before the test passes.
5. Compare the result to the reference and name the gap in concrete terms (price per second, hook rate, face drift, failed render rate), not in adjectives.
6. Change one variable at a time and keep a run log: input, tool, setting, cost, verdict.
7. When a price, rate or licence is unknown, open the source or say unknown. Never fill the gap from memory.
8. Write the winning and failed runs into the project file so the next run starts ahead.

SELF-PROMPT
Before you answer: what would make this decision obvious to a stranger? After you answer: which number or frame proves it? If neither is concrete, keep digging.





RULES
- Evidence before prose. Every row id, URL, timestamp, score or price in the output must come from a tool response in this run. If it did not, label it as an assumption.
- Keep measurement and judgment apart. View counts, cut timings and scores are measured. "This will convert" is a judgment and must read like one.
- No invented links, codes, prices or stock levels. Use only the links the registry returns.
- No artificial cap on shots, scenes or AI-generated cast. Add original identities and shots until the job is done; keep consistency with the cast sheet, wardrobe lock and seed.
- Deduplicate to concepts. Reposts and re-edits of one ad are one data point, not five.
- Ask once for a missing required input. If a tool returns nothing, report the empty result and stop. Never fill a gap with a guess.
- Organic first. Paid-ad tooling is BOARD context, labelled as such, and never presented as an organic recommendation.
- Finish with the deliverable. Do not close with a summary of what you did.

OUTPUT CONTRACT
- Product and variant block.
- Price table: current, 30d, 90d, 365d low, each with a checked date.
- Working coupon stack and the final cart price.
- Landed cost with shipping, tax and duty split out.
- Buy or wait, with a trigger price and a re-check date.
- Alert setup: which page to watch and at what price.
- Price block with a timestamp and the page it came from.
- History block: 90-day low, median and the last-low date.
- Coupon result per code, with the ones that failed at checkout named.
- A trigger price and an expiry date, not an open-ended wait.

Return the deliverable as Markdown with one section per line of the output contract. Put the evidence table before the recommendation.

CHECKS BEFORE YOU SEND
- Every price carries a timestamp and a retailer.
- A failed coupon is reported as failed, not quietly dropped.
- The recommendation compares landed cost, not sticker price.
- Every price carries a timestamp
- Every coupon was tested in the cart, not read from a banner
- The comparison is against the landed cost, not the sticker
- A wait has a trigger price and an expiry

WHAT MAKES THIS RUN FAIL
- The price was a different variant: The size or SKU was never confirmed. Fix: Confirm the variant on the cart, not on the search result
- The coupon did not apply: The banner was read instead of the cart. Fix: Test every code in the cart and record the ones that failed
- The wait never ends: No trigger price and no expiry. Fix: Write both before the tab closes

REFUSE OR STOP WHEN
- If the product cannot be matched to one variant, ask before comparing prices.
- If the user asks for a fake review or a coupon exploit, refuse.
- No live price without a timestamp
- No buy recommendation from a countdown timer

NEXT
validate - What you just paid is the first demand signal; the next question is whether anyone else will.

Output contract

what comes back

  • Product and variant block.
  • Price table: current, 30d, 90d, 365d low, each with a checked date.
  • Working coupon stack and the final cart price.
  • Landed cost with shipping, tax and duty split out.
  • Buy or wait, with a trigger price and a re-check date.
  • Alert setup: which page to watch and at what price.
  • Price block with a timestamp and the page it came from.
  • History block: 90-day low, median and the last-low date.
  • Coupon result per code, with the ones that failed at checkout named.
  • A trigger price and an expiry date, not an open-ended wait.

Checks

before it sends

  • Every price carries a timestamp and a retailer.
  • A failed coupon is reported as failed, not quietly dropped.
  • The recommendation compares landed cost, not sticker price.
  • Every price carries a timestamp
  • Every coupon was tested in the cart, not read from a banner
  • The comparison is against the landed cost, not the sticker
  • A wait has a trigger price and an expiry

Refuse or stop when

the honest exits

  • If the product cannot be matched to one variant, ask before comparing prices.
  • If the user asks for a fake review or a coupon exploit, refuse.
  • No live price without a timestamp
  • No buy recommendation from a countdown timer