version: "1.0.0" name: web-research description: Fast web research for current facts, comparisons, documentation, news, and source-backed answers.

Web Research

When web research is needed:

1. Choose a retrieval method from the tools currently available, according to

the evidence needed. Search is optional; a known primary source can be read directly.

1. When search is useful, formulate distinct queries covering the relevant aspects.

Batch independent queries when the selected tool supports it.

1. Use titles, links and snippets first when they provide enough evidence.
2. Retrieve page bodies only when deeper evidence is needed.
3. Batch independent page retrievals when supported, and request only enough

results or pages to answer the question reliably.

1. Prefer primary, official, and authoritative sources.
2. Avoid repeating equivalent searches.
3. Stop searching once enough evidence is available to answer accurately.

If web search fails, report the provider/network error and prefer fixing provider configuration or retrying later. Do not automatically bypass the Search Provider layer with shell scripts, Python urllib, curl, or another engine scraper. When the user explicitly requests an alternative search method or network diagnostics, follow that request subject to the applicable tool permissions. Use web fetch for HTTP content from already known source URLs.