nesdeq/openclaw-serper
Overview
This skill performs Google searches via the Serper API and returns full, readable page text for each result using trafilatura. It delivers cleaned article or page content — not just snippets — so you get the original, extractable text for research or time-sensitive queries. Use it when up-to-date, sourceable web content is required.
How this skill works
You supply one well-crafted query and the skill runs a single Google search (two modes available: default for all-time results and current for recent/news). For each result the tool fetches the page and extracts the full readable text with trafilatura, streaming a JSON array that includes metadata plus full content for each result. The output already contains the page text, so no additional URL fetching is necessary or allowed.
When to use it
- General research, tutorials, technical documentation, and product comparisons
- News, recent developments, breaking events, or anything time-sensitive (use current mode)
- Fact checks and citations where full article context matters
- When you need cleaned, readable full-page text rather than short search snippets
- Comparisons across multiple sources in a single response
Best practices
- Craft one clear, focused search query — one good query is usually enough
- Choose mode: default for evergreen topics, current for past-week/news + Google News
- If the query targets a specific country or non-English language, set locale flags (--gl and --hl) accordingly
- Do not run multiple sequential searches to “dig deeper”; use the full page texts returned
- Limit to at most two searches only when the request genuinely spans two distinct topics
Example use cases
- Compare technical guides and extract step-by-step instructions across multiple sources
- Gather full news articles and press releases for a recent product announcement
- Pull complete product pages and reviews for a buying comparison
- Collect source material and quotes for a research summary or report
- Retrieve full documentation pages when exact wording and context matter
FAQ
default returns 5 web results with full page text; current returns 3 web + 3 news results from the past week, each with extracted content.
Can I run multiple searches to cover different angles?
Generally no — one well-crafted query gives broad coverage. Use at most two searches only when the request truly involves two distinct topics.
What if a page can’t be extracted?
If trafilatura fails or the site blocks scraping, the result falls back to the search snippet provided by Serper.
