tavily-dynamic-search

Runs web searches and extracts evidence via local Python filtering.

Updated May 3, 2026
One-click install
npx skills add https://github.com/LuisElPoti/launch-signal-dashboard --skill tavily-dynamic-search-luiselpoti
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tavily-dynamic-search
Source: https://github.com/LuisElPoti/launch-signal-dashboard/tree/main/.agents/skills/tavily-dynamic-search
Command: npx skills add https://github.com/LuisElPoti/launch-signal-dashboard --skill tavily-dynamic-search-luiselpoti

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Web research often floods the context window with irrelevant HTML, navigation, and boilerplate, making it hard to extract the facts you actually need.

Core Features & Use Cases

  • Programmatic web search: Run targeted searches and triage results by score, title, URL, and snippet.
  • Context-isolated extraction: Use local Python to filter and print only the relevant parts, so raw page content does not enter your context window.
  • Multi-turn research workflow: Save raw search results to disk for later turns, then extract only the sections that match your query.
  • Use Cases: Gathering up-to-date product/news details, pulling specific facts (dates, metrics, pricing, quotes), and building curated evidence for analysis or outreach.

Quick Start

Ask the agent: research recent funding announcements for a company and return only the relevant dates, amounts, and source links after filtering out boilerplate.

Frequently Asked Questions about tavily-dynamic-search

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract specific facts from web search results without flooding my context window?

You can isolate web search context by using local Python filtering to process raw page content, printing only the curated facts, metrics, and source links so HTML boilerplate never enters the model context.

What is context-isolated web research and how does it work?

Context-isolated web research performs programmatic searches and triages results by score and snippet locally, preventing raw page content from entering the model context by emitting only curated print() outputs after applying deterministic filtering.

Can I save web search results to disk for multi-turn research workflows?

Yes, you can save raw search results to disk for later turns, then extract only the sections matching your query across multi-turn refinement workflows to build curated evidence for analysis or outreach.

Does Tavily support programmatic web search and evidence gathering with JSON output?

Yes, Tavily supports search and extraction via tvly with JSON output, allowing you to gather up-to-date product details, news, pricing, and quotes while applying deterministic selection logic and error handling locally.

What are the limitations of using local Python filtering for web research?

Local Python filtering for web research requires deterministic selection logic to handle errors and triage results by score, title, URL, and snippet, meaning raw page content bypasses the model context entirely, limiting direct interaction with full HTML.