菜单
Produces a structured Research Summary document (findings, sources, attributions) from targeted web queries using Brave Search and web_fetch. Use when: 'research this online', 'find current information about X', 'verify this claim', 'gather sources on Y', 'search for what the field says about Z'.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/DojoGenesis/gateway --skill web-research命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | web-research |
| model | sonnet |
| description | Produces a structured Research Summary document (findings, sources, attributions) from targeted web queries using Brave Search and web_fetch. Use when: 'research this online', 'find current information about X', 'verify this claim', 'gather sources on Y', 'search for what the field says about Z'. |
| category | continuous-learning |
| triggers | ["research this online","find current information about","verify this claim","gather sources on","search for what the field says about"] |
| tier | 1 |
| agents | ["primary"] |
| tool_dependencies | ["web_tools"] |
| inputs | [{"name":"query","type":"string","description":"The research question or topic to investigate online","required":true}] |
| outputs | [{"name":"research_summary","type":"ref","format":"cas-ref","description":"Structured Research Summary document with findings, sources, and attributions from web queries"}] |
Version: 1.1 Created: 2026-02-02 Author: Cipher (self-taught) Purpose: Effective web research using Brave Search API and web_fetch for content extraction
This skill encodes best practices for web research — finding, evaluating, and synthesizing information from online sources. It provides patterns for search query formulation, source evaluation, information synthesis, and attribution. Web research is not about collecting links — it's about building understanding from reliable sources.
Output templates for all research modes are in references/research-output-templates.md.
Available Tools:
web_search(query, count, country, search_lang, ui_lang, freshness) — Brave Search API for finding sourcesweb_fetch(url, extractMode, maxChars) — Extract readable content from URLs (markdown/text)Parameters:
query: Search query string (required)count: Number of results (1-10, default: 5)country: 2-letter country code for regional results (default: 'US')search_lang: ISO language code for search resultsui_lang: ISO language code for UI elementsfreshness: Time filter for results (pd=past 24h, pw=past week, pm=past month, py=past year)url: HTTP or HTTPS URL to fetchextractMode: 'markdown' or 'text' (default: 'markdown')maxChars: Maximum characters to return (truncates when exceeded)Before searching, clarify:
Examples:
Pattern: [topic] [context/aspect] [specific keywords] [optional: date filter]
Examples:
| Topic | Bad Query | Good Query |
|---|---|---|
| AROMA | "AROMA" | "AROMA agent collaboration v0.0.1 specification" |
| Next.js bugs | "Next.js bugs" | "Next.js 16.1.6 security vulnerabilities CVE" |
| AI research | "AI research" | "AI agent memory compression techniques 2025" |
| Excel design | "Excel design" | "Excel data visualization best practices conditional formatting" |
Freshness filters:
freshness="pd" (past 24 hours)freshness="pw" (past week)freshness="pm" (past month)freshness="py" (past year) or no filterExecute search:
web_search(
query="your query here",
count=5,
country="US",
search_lang="en"
)
Evaluate each source:
| Criterion | What to Check | Good Signs | Bad Signs |
|---|---|---|---|
| Credibility | Domain reputation | .gov, .edu, established news | Unknown blog, social posts |
| Relevance | Title and snippet match | Direct answer to question | Tangential content |
| Freshness | Publication date | Recent (past 1-3 months) | Outdated (>1 year old) |
| Depth | Content length and detail | 500+ words, specific details | 200-word overview |
| Authority | Author or org expertise | Named experts, official docs | Anonymous, generic content |
Source selection:
For relevant sources:
web_fetch(
url="https://example.com/page",
extractMode="markdown",
maxChars=20000
)
Extraction strategy:
Goal: Answer the research question, not regurgitate content.
Synthesis template — see references/research-output-templates.md for the full Research Synthesis format.
Synthesis principles:
Use when: Verifying a specific claim, fact, or data point
Process:
Output: Verification document with Verdict (Confirmed / False / Partially Confirmed / Could Not Verify), Evidence section listing each source's position — see references/research-output-templates.md
Use when: Need comprehensive understanding of a complex topic
Process:
Output: Deep Dive document organized by subtopics with cross-subtopic synthesis — see references/research-output-templates.md
Use when: Broad pattern recognition across many topics
Process:
Output: Landscape Scan document with themes, key players table, and gaps list — see references/research-output-templates.md
Before concluding web research, verify:
freshness filterUse with:
research-modes — For structuring deep vs. wide research phasesspecification-writer — When research feeds into spec writingseed-extraction — When research reveals reusable patternsworkspace-navigation — When organizing research findings in shared workspacesPattern:
web_search() -> web_fetch() -> Extract insights ->
research-modes(structure) OR specification-writer(draft) OR seed-extraction(capture)
docs/research/ or scouts/ directory[date]_[topic]_web_research.mdScenario 1: "Find current information about WebSocket performance benchmarks for Go" → Verification/Deep Dive document with 3-5 authoritative sources, specific numbers, attribution per claim, and open questions flagged
Scenario 2: "Search for what the field says about AI agent orchestration patterns" → Landscape Scan document with theme clusters, key tool/framework table, and identified gaps across 8-10 sources
Related Skills:
research-modes — Deep vs. wide research structuringspecification-writer — Research to spec conversionseed-extraction — Pattern extraction from research findingsLast Updated: 2026-04-08 Maintained By: Cipher Status: Active