← Back to Skills Marketplace
932
Downloads
0
Stars
4
Active Installs
1
Versions
Install in OpenClaw
/install moss-deep-search
Description
Deep web research using LLM + multiple search rounds. Use when user needs comprehensive analysis, detailed research, or says "deep search", "详细搜索", "深入调查".
README (SKILL.md)
Deep Search Skill
使用 LLM 进行深度研究，通过多轮搜索和综合分析获取详细信息。
触发条件
- 用户说 "deep search"、"深度搜索"、"详细调查"
- 需要多来源综合分析
- Brave 搜索结果不够深入
工作流程
- 分析查询，制定搜索策略
- 执行多轮搜索（3-5 轮）
- 整合信息，生成综合报告
- 发送到指定渠道
输出格式
- 执行计划（搜索轮次和关键词）
- 分章节的综合报告
- 信息来源链接
- 总结和建议
Usage Guidance
This is an instruction-only 'deep search' skill that generally matches its stated purpose, but it is vague about how searches are performed and where results are sent. Before installing or using it: (1) ask the developer which search/browsing tools or APIs it will use (and whether the agent will access the web interactively), (2) ask what '指定渠道' (designated channels) means and whether any credentials or endpoints will be configured or required, (3) avoid sending sensitive data through it until you confirm destination and retention policies, and (4) prefer a version that explicitly lists allowed search sources and requires explicit user confirmation before transmitting results externally.
Capability Analysis
Type: OpenClaw Skill
Name: moss-deep-search
Version: 1.0.0
The skill bundle contains standard metadata and instructions for a multi-round search and analysis workflow. The instructions in SKILL.md are consistent with the stated purpose of performing deep research and do not contain any malicious commands, data exfiltration logic, or prompt injection attacks.
Capability Assessment
Purpose & Capability
Name and description (deep web / multi-round search) align with the SKILL.md workflow (analyze query, run multiple search rounds, synthesize report). The skill does not request unrelated binaries, env vars, or installs.
Instruction Scope
Instructions are high-level and allow broad agent discretion: they require '执行多轮搜索' and to '发送到指定渠道' (send to specified channels) but do not define which search sources, browsing tool, or channels/credentials. This vagueness could lead the agent to transmit results to external endpoints or use unapproved browsing/scraping behaviors.
Install Mechanism
Instruction-only skill with no install spec or code files — minimal disk or execution footprint from the skill itself.
Credentials
No environment variables or credentials are requested (proportionate), but the SKILL.md's instruction to 'send to specified channels' implies outbound delivery that would normally require endpoints/credentials — the absence of declared channels/creds is inconsistent and unclear.
Persistence & Privilege
always:false and no install/runtime behavior that requests persistent presence or modifies other skills or system settings.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install moss-deep-search - After installation, invoke the skill by name or use
/moss-deep-search - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of moss-deep-search (deep-search) skill.
- Enables deep web research using multiple search rounds and LLM analysis.
- Triggered by requests for comprehensive, detailed search or analysis (e.g., "deep search", "详细搜索", "深入调查").
- Analyzes queries, iteratively searches (3–5 rounds), aggregates and reports findings.
- Outputs include: execution plan, chaptered report, source links, and actionable summary.
Metadata
Frequently Asked Questions
What is Moss Deep Search?
Deep web research using LLM + multiple search rounds. Use when user needs comprehensive analysis, detailed research, or says "deep search", "详细搜索", "深入调查". It is an AI Agent Skill for Claude Code / OpenClaw, with 932 downloads so far.
How do I install Moss Deep Search?
Run "/install moss-deep-search" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Moss Deep Search free?
Yes, Moss Deep Search is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Moss Deep Search support?
Moss Deep Search is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Moss Deep Search?
It is built and maintained by jiangwzh (@jiangwzh); the current version is v1.0.0.
More Skills