⚠️ This column is a free educational sharing of the prompt creation process, and is not any investment, legal, or financial advice. AI execution results may contain errors and independent verification is essential, and all responsibility for usage is entirely attributed to the user themselves. It cannot be used directly for business decision-making. [View all mandatory confirmation items below the 30-week timeline before use[]](https://www.notion.so/30-29f84ddcec3c80a1a87eda9883d81453?pvs=21)
In Week 8, we completed the first version of Product_Checklist.md. It was about 1,000 lines, with 450 lines of rules in PART A and 550 lines of source-specific strategies in PART B. We referenced the Strategy_Checklist.md structure and added the insert command and original preservation concept.
Now it was time to actually start searching. We requested Claude, "Read Product_Checklist.md and start searching." Claude read the checklist with view and identified the first task with Sequential Thinking. "I will start searching the first source, Proposed Rule.”
It searched with web_search and results appeared. 2-3 documents appeared and Claude showed a summary. "This document is a Proposed Rule. It contains semiconductor-related content." But there was a problem.
Looking at the search results, only summaries appeared. "This document proposes tariffs on semiconductor Memory chips, Processors, detailed product list is in the appendix.", we knew there was a list in the appendix, but the problem was we couldn't see the entire list.
web_search was getting a summary of the search results. It wasn't the entire content of the document, only the title, date, and brief description. It said "60 product list in appendix" but we couldn't know what the 60 items were.
Eventually we had to check the entire document. We had to look directly at the appendix to see what was there besides Memory chips, what each product description was, and what the HTS codes were. It was impossible with web_search.
That's when we had to use web_fetch. Unlike web_search, it's a command that brings most of the web page content without summarizing, bringing all text content of the page including the appendix. Table formats have HTML tags removed but the data comes as is, so we could see the entire product list. However, since it consumes many tokens at once, it couldn't be used for all documents, so it had to be used selectively.