#AISChallenge - Day 3 · AI Automation Society

#AISChallenge - Day 3

What I built: /research — buyer keyword extractor with HTML gallery output

The skill searches the web via Firecrawl, scrapes the top results, and produces a self-contained HTML report you can open directly in a browser. It extracts product keywords organized by category (synonyms, product names, specs, use cases, navigation labels) + an image gallery pulled from every <img> tag on the page.

Trigger: /research <topic> — or just say "research X for the UK market" and it picks it up naturally.

One optimization I made:

The first version returned a markdown text report. After watching it run, I realized the output was all words and no visual context — I had no feel for how the product actually looks on pages buyers visit. So I added formats: ["html"] to the Firecrawl call, extracted image URLs from <img src> tags, and rewrote the output as a styled HTML file with a product gallery. Now every report shows the actual photos customers see alongside the keywords.

Second smaller one: I added a capital city rule, it automatically appends "Capital city" to the search query. If you say "German market" it appends "Berlin". This anchors results to where buyers in that country actually search from, instead of getting generic global results.

Skill is ~130 lines. Built iteratively in one session by running real searches and adjusting based on what came back.

0 comments