tavily-summary
AI-optimized web search with structured summarization, combining Tavily Search API + proven summarization methodology from summarize.
- License 许可
- MIT
- 最近更新
- 2026-06-10
- Stars
- ★ 3
- Forks
- 0
01 前言 Front Matter
- name
- tavily-summary
- homepage
- https://tavily.com
- metadata
- { "clawdbot": { "emoji": "🔍", "requires": { "env": [ "TAVILY_API_KEY" ], "bins": [ "node" ] }, "primaryEnv": "TAVILY_API_KEY" } }
- description
- AI-optimized web search with structured summarization, combining Tavily Search API + proven summarization methodology from summarize.
03 SKILL.md 正文
Tavily Summary
AI-optimized web search with structured summarization methodology, combining:
- Tavily Search API for clean, relevant search results (AI-native)
- Proven summarization rules extracted from the
summarizeproject - Output follows consistent formatting rules for readability
Features
- Search the web with AI optimization (better than plain search engines for AI agents)
- Automatically summarizes results following structured formatting rules:
- Strict length control, no padding
- Automatic heading hierarchy for long content
- Proper citation of key excerpts
- Ad/sponsor content automatically stripped
- Consistent Markdown formatting
Usage
Search
node {baseDir}/scripts/search.mjs "query"
node {baseDir}/scripts/search.mjs "query" -n 10
node {baseDir}/scripts/search.mjs "query" --deep
node {baseDir}/scripts/search.mjs "query" --topic news
node {baseDir}/scripts/search.mjs "query" --topic news --days 30
Options
-n <count>: Number of results (default: 5, max: 20)--deep: Use advanced search for deeper research (slower, more comprehensive)--topic <topic>: Search topic -general(default) ornews--days <n>: For news topic, limit to last n days
Extract content from URL
node {baseDir}/scripts/extract.mjs "https://example.com/article"
Output Summary Rules
Core Principles
- Positioning: Extract key points for users, let them quickly grasp core info and decide whether to read in depth
- Authenticity: 100% based on search results, never fabricate information
- Conciseness: Strictly control length, prefer shorter, never pad for no reason
- Formatting: Use Markdown formatting, long content must split into sections with headings
Formatting Rules
- Content longer than 500 characters equivalent → split with
###prefixed Markdown headings, at least 3 headings - Short paragraphs, use bullet points only when improves readability, don't force rigid templates
- Include 1-2 short exact excerpts (max 25 words each) as italicized, ignore ads/sponsors/promotions completely, treat them as if they don't exist, don't mention ignoring them
Output Structure (for news search example)
- One sentence opening summary of the core conclusion
- Break into points/events, each with core idea + key data
- (Optional) 1-2 italicized original excerpts
- One sentence closing summary
Absolute Prohibitions
- ❌ Never include ads/sponsors/promotions/CTAs, delete directly
- ❌ Use straight quotes only, no curly quotes
- ❌ Don't add emojis, disclaimers, unfounded speculation
- ❌ Never exceed requested length, finish early if content is short, don't pad
- ❌ Don't fabricate excerpts, omit if no suitable excerpt exists
Requirements
TAVILY_API_KEYfrom https://tavily.com (free plan available: 1,000 searches/month)- Node.js 18+
Credits
04 同仓库更多 Skills
youtube-content
YouTube transcripts to summaries, threads, blogs.
xurl
X/Twitter via xurl CLI: post, search, DM, media, v2 API.
touchdesigner-mcp
Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.
test-driven-development
TDD: enforce RED-GREEN-REFACTOR, tests before code.
teams-meeting-pipeline
Operate the Teams meeting summary pipeline via Hermes CLI — summarize meetings, inspect pipeline status, replay jobs, manage Microsoft Graph subscriptions.
systematic-debugging
4-phase root cause debugging: understand bugs before fixing.