HiAi-gg Research the Web with Firecrawl
Use this skill when an agent needs current, web-sourced information — the Firecrawl equivalent of "look it up online", with citation discipline.
When to use
- Questions whose answer changes over time (docs, pricing, features, releases).
- Competitive/ecosystem research across multiple sources.
- Verifying claims against the current web.
- Do not use for stable, self-contained knowledge the agent already has.
Workflow
1. Plan the queries
- Break the research question into 2–4 concrete search queries (distinct phrasings, competitors, or aspects).
- Note which results you will treat as authoritative (vendor docs, official sources) vs secondary.
2. Search
firecrawl_search(query, limit)
Run each planned query. Record URLs, titles, and descriptions for the top results. On the keyless tier the free rate limit applies — batch queries instead of firing one per second.
3. Read the important pages
For the 1–3 most authoritative results, read the page content:
firecrawl_scrape(url)
Extract the facts relevant to the question. Prefer official/vendor sources for facts about a product.
4. Synthesize with citations
Answer the question from the gathered material. Cite each claim to its source URL. Distinguish:
- facts from official sources,
- claims from secondary sources,
- gaps ("no current source found for X").
5. Report
Give: the answer, the supporting sources (title + URL per claim), and anything you could not verify. Do not fabricate citations — only cite pages you actually searched/scraped.
Guardrails
- Cite real sources; never invent URLs.
- Respect the keyless rate limit; batch queries.
- Do not pass API keys through tool arguments.