Create a complete order for the selection of goods with a price comparison. Return JSON ONLY: {"goal":"what to choose","criteria":"required and desirable conditions with priority","compare_by":"by which columns to compare","output":"result type, sorting, number of lines, links","limits":"borders: how many stores, region, no payments"}
Here for the first time you combine the goal, criteria, format and boundaries into one work assignment. The selection task with prices is a classic for an agent, because there is a lot of manual routine: go around the stores, compare prices, weed out the unsuitable. Mechanics: the agent goes through the sites, collects the same set of fields for each model and puts them into a table - but only if you specified this set. Technique of the masters: distinguish between “model” and “sentence”. One model is sold in different stores at different prices - ask for the "store" and "price" columns to see the best offer for each model, rather than one random price. The second trick: add a “price check date” - prices do not last long, and without a date the code silently becomes obsolete. Insider: set the “reviews” criterion in a measurable way – “a rating of at least 4.3 and at least 50 reviews”, otherwise the agent will choose a nice number without volume. A common mistake is to ask for “top 5 best” without axes of comparison. Then the “best” will simply be the first to be caught, and the decision will have to be made blindly.
Unlock access to submit solutions for instant AI review.