Learning Timeline
Key Insights
Token-Saving Strategy (Game Plan)
Always ask Claude for a 'game plan' before starting heavy tasks. This allows you to correct the AI's logic before it starts burning through your API tokens on potentially incorrect data.
Avoid 'Massive Laundry Lists'
Don't provide oversized CSV lists or request too many data points (images, pricing, amenities, etc.) at once. AI tends to make mistakes when the workload is too complex. It’s better to process it in stages.
The Importance of Data Refinement
Data enrichment is rarely perfect on the first pass. You may need to perform 2-3 reruns to ensure all edge cases are handled and every row of data (such as stall information) is fully complete.
Prompts
Inventory Enrichment Prompt
Target:
Claude Code
Go and find all the different types of luxury restroom trailers that this business offers. Look for specific details like the number of stalls for each trailer. Give me your game plan and tell me if I'm missing anything before we go for it.
Step by Step
Data Enrichment Workflow Using Claude Code & Crawl4AI
- Identify the list of URLs or business websites you want to analyze (e.g., a list of luxury restroom trailer suppliers).
- Open your terminal or Claude Code workspace.
- Integrate Crawl4AI into the workflow to allow Claude to browse the target websites.
- Enter a specific prompt to find detailed product information (such as the number of stalls for each trailer model).
- Add the command 'Give me your game plan' at the end of the prompt to ask Claude to explain its intended steps.
- Verify the game plan provided by Claude to avoid wasting tokens and ensure the search logic is correct.
- Run the crawling process and let Claude extract the data into your desired format (such as a list of trailer types).
- Review the output to identify any edge cases or incomplete data after each pass.
- Perform a 'rerun' or restart the enrichment process 2 to 3 times with more refined instructions if any information is still missing.
