Claude to compare speakers from local catalogs
Daniel Rosehill published a Claude Code workflow for choosing a portable Bluetooth speaker. He supplied a specification file describing the intended use and budget, alongside screenshots from retailer catalogs, so the comparison could start from available products rather than an unconstrained product search.
github.com
Original prompt
Shared by the source author. See source notes for context.
/recommend
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Daniel Rosehill
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The repository contains a structured speaker comparison and PDF recommendation, including a JBL Flip 7 offer at a reported ₪349.
- Human involvement
- Rosehill supplied personal specifications and catalog captures; the source does not describe autonomous payment or a completed purchase.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Write the intended speaker use and budget in a specification file.
- Extract catalog screenshots into JSON and research the shortlisted products.
- Run the recommendation command and generate a Typst PDF with prices.
Daniel Rosehill published a Claude Code workflow for choosing a portable Bluetooth speaker. He supplied a specification file describing the intended use and budget, alongside screenshots from retailer catalogs, so the comparison could start from available products rather than an unconstrained product search.
The repository separates catalog extraction from price analysis. A catalog scanner turns the captured offers into structured JSON, and a price analyst compares the candidates with the personal requirements. The workflow then produces a recommendation report in Typst and exports it as a PDF.
Rosehill used Gemini through MCP for research grounding. The repository also discusses browser and scraping configurations, but those optional integrations are not evidence that every one was used for this comparison. The reported recommendation was a JBL Flip 7 listed at ₪349 by KSP, with other lower priced and premium candidates retained in the comparison.
This is a product evaluation artifact, rather than a verified purchase. The repository includes estimates about sound and relative value, but no measured loudness test or order receipt. The practical pattern is to keep the user's requirements and the extracted retailer offers separate, then ask the agent to explain a recommendation from those two inputs. Rosehill published the workflow and the generated report for inspection.
Prompt
Original prompt
Source: github.com · Daniel Rosehill
Frequently asked questions
What prevented an unconstrained product search?
A specification file supplied the personal use and budget, while captured retailer catalogs supplied candidate offers.
Which research integration was actually reported?
Gemini through MCP supplied research grounding; other browser configurations were optional.
Was the recommended speaker purchased?
The source publishes the evaluation and report, without an order receipt or verified purchase.
More details
About this account
Claude Code scanned catalog offers, structured product data and generated a recommendation report using research from Gemini MCP.
Firsthand source opened and read. Prompt is a verbatim excerpt of the published original instruction.
Credit: Daniel Rosehill
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Gemini MCP · Typst · GitHub
Speaker comparison · Catalog screenshots · Local retailers · Typst PDF
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.