← 返回 Skills 市场
openlark

Search Synthesis Expert

作者 OpenLark · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ 安全检测通过
51
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install search-synthesis-expert
功能描述
Uses sequential-thinking to decompose tasks and formulate search strategies, browser automation (Playwright) for multi-source information search and collecti...
使用说明 (SKILL.md)

Search Synthesis Expert

Expert at searching and synthesizing information. Executes tasks in a three-phase process: Plan → Collect → Review.

Workflow

User query → Phase 1: Decompose & Plan → Phase 2: Browser search & collect → Phase 3: Review & Synthesize → Output report

Phase 1: Decompose & Plan (sequential-thinking)

Use step-by-step reasoning to decompose tasks and formulate search strategies:

  1. Clarify goal: Understand the user's core question and information boundaries
  2. Decompose sub-problems: Break complex questions into independently searchable sub-questions
  3. Formulate search strategy: Determine keywords, target sites, and search order for each sub-problem
  4. Prioritize: Order searches by information importance

Example output:

Task decomposition:
1. Sub-question A → Search keywords: {keywords} → Target: {site}
2. Sub-question B → Search keywords: {keywords} → Target: {site}
3. Sub-question C → Search keywords: {keywords} → Target: {site}

Phase 2: Browser Search & Collect (Playwright)

Use Playwright to simulate user browsing for search and information collection:

  1. Open search engine or target site
  2. Enter search keywords, simulating real user behavior
  3. Browse search results, click relevant links
  4. Extract page content, collect key information
  5. Multi-source comparison: Get information from different sources for the same question
  6. Record sources: Save each piece of information's URL for traceability

Core operations:

# Example operation flow (pseudocode)
page.goto("https://www.google.com")
page.fill("input[name='q']", keyword)
page.press("input[name='q']", "Enter")
page.wait_for_selector("div#search")
results = page.query_selector_all("h3")
# Click relevant results → Extract content → Record sources

Phase 3: Review & Synthesize (sequential-thinking)

Use step-by-step reasoning to review collected information:

  1. Deduplicate: Merge duplicate information from multiple sources
  2. Cross-validate: Compare consistency across sources, flag contradictions
  3. Credibility assessment: Evaluate information credibility based on source authority
  4. Structured synthesis: Organize final output by logical relationships
  5. Gap marking: Flag information gaps that could not be filled

Output Format

═══════════════════════════════════
Search & Synthesis Report

📋 Query
{original user question}

🔍 Search Strategy
{sub-problems and keyword list}

📚 Sources
| # | Source | Summary | Credibility |
|---|--------|---------|-------------|
| 1 | {URL} | {summary} | ⭐⭐⭐ |
| 2 | {URL} | {summary} | ⭐⭐ |

📝 Synthesized Conclusion
{structured complete answer}

⚠️ Notes
- {information limitations}
- {uncovered aspects}
- {suggestions for further research}
═══════════════════════════════════

Best Practices

Principle Description
Multi-source verification Each key fact verified from at least 2 independent sources
Source traceability Each piece of information tagged with source URL for backtracking
Timeliness first Prioritize latest information, note information date
Breadth first Search same question from different angles (official docs, communities, blogs, news)
Honest labeling Clearly mark unfound information as "not found," never fabricate

Notes

  • Set reasonable User-Agent and delays when using Playwright to avoid anti-bot measures
  • For sites requiring login, mark as "login required"
  • Non-deterministic information (predictions, opinions) must note the source's position
  • If search cannot cover all sub-problems, mark missing parts in the report
安全使用建议
Install this only if you are comfortable with an agent using browser automation to send search queries and visit third-party websites. Avoid giving it confidential prompts, internal project names, credentials, or sensitive personal data unless you are comfortable with those terms appearing in search engines or target sites.
能力评估
Purpose & Capability
The stated purpose is multi-source search and synthesis, and the Playwright browsing instructions directly support that purpose.
Instruction Scope
The workflow is scoped to user-directed research, source collection, and synthesis; it does not include prompt overrides, hidden actions, destructive steps, or unrelated data access.
Install Mechanism
The package contains only a single markdown skill file, with no executable scripts, dependencies, install hooks, or declared package installs.
Credentials
Live browser searches and opening third-party pages are proportionate for this skill, but users should understand that search terms and visited pages may be exposed to external sites.
Persistence & Privilege
No persistence, background workers, privilege escalation, credential harvesting, local profile use, or session reuse is instructed; sites requiring login are to be marked as login required.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install search-synthesis-expert
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /search-synthesis-expert 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release of search-synthesis-expert skill. - Introduces a structured three-phase workflow: Plan (decompose & strategize), Collect (browser-based search via Playwright), Review (synthesize & validate). - Automates multi-source information gathering and synthesis using browser automation for deep research, technical investigation, and multi-source fact-checking. - Outputs a comprehensive report including query, search strategy, source summaries, synthesized conclusions, and notes on limitations and gaps. - Emphasizes principles of multi-source verification, source traceability, credibility assessment, and transparent disclosure of limitations.
元数据
Slug search-synthesis-expert
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Search Synthesis Expert 是什么?

Uses sequential-thinking to decompose tasks and formulate search strategies, browser automation (Playwright) for multi-source information search and collecti... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 51 次。

如何安装 Search Synthesis Expert?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install search-synthesis-expert」即可一键安装,无需额外配置。

Search Synthesis Expert 是免费的吗?

是的,Search Synthesis Expert 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Search Synthesis Expert 支持哪些平台?

Search Synthesis Expert 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Search Synthesis Expert?

由 OpenLark(@openlark)开发并维护,当前版本 v1.0.0。

💬 留言讨论