← Back to Skills Marketplace
Kimi Websearch
by
henryzhuhr
· GitHub ↗
· v1.0.9
· MIT-0
591
Downloads
0
Stars
2
Active Installs
9
Versions
Install in OpenClaw
/install kimi-websearch
Description
联网搜索工具。Use When (1)用户需要通过联网获取信息; (2)如果你无法直接回答用户的问题，或者需要更多信息来回答用户的问题。
README (SKILL.md)
Kimi Web Search / Kimi 联网搜索工具
Requirements
通过设置环境变量 KIMI_API_KEY 或 MOONSHOT_API_KEY 来提供 Kimi/Moonshot API 密钥。
如果没有设置，请登录 Moonshot API Keys 获取 API Key，并将其设置为环境变量。
执行流程
脚本说明
⚠️ 注意: 如果当前环境使用 uv ，则优先使用
uv run xxx.py来执行脚本；如果没有 uv ，则使用python3 xxx.py来执行脚本。
联网搜索脚本
# 通过 Kimi API 进行联网搜索，获取最新的网络信息来回答用户的问题。
python3 {baseDir}/scripts/web_search.py "\x3Cthe question you want to ask>"
# 当遇到错误的时候，可以获取帮助
python3 {baseDir}/scripts/web_search.py --help
输出要求
-
必须提供给用户消息来源的url，确保信息的准确性，例如
1. 新闻内容xxxx 消息来源：[xxx](https://www.example.com) -
如果没有明确的搜索主题，例如“今天的新闻”，可以参考以下分类整理后输出给用户：
📰 今日要闻
🇨🇳 国内
🌍 国际焦点
📈 财经资讯
💻 科技动态
- 优先输出整理后的结论，不要把整段原始脚本输出原封不动贴给用户。
- 如果搜索结果明显不完整、存在时效风险，直接说明不足。
- 如果用户要求继续追问同一主题，直接换一个搜索参数重新运行。
Usage Guidance
This skill uses your Moonshot/Kimi API key and sends your prompts/conversation to https://api.moonshot.cn to obtain search results — if you install it, be aware your queries and context will go to that service (and may incur usage/costs). The included Python script itself does not crawl the web locally; it relies on Moonshot's model/tooling ($web_search) to return results and URLs. If you need private/offline searching or do not trust the Moonshot endpoint, do not provide your API key. Otherwise, the requested env vars and the code appear proportionate to the described websearch purpose. Verify you trust the Moonshot provider and that the API key you supply has appropriate scope/permissions and billing limits.
Capability Analysis
Type: OpenClaw Skill
Name: kimi-websearch
Version: 1.0.9
The skill provides a legitimate interface for Kimi/Moonshot web search but is classified as suspicious due to a potential shell injection vulnerability. In SKILL.md, the agent is instructed to execute `web_search.py` by passing the user's query directly into a shell command within double quotes, which could lead to arbitrary command execution if the input is not properly sanitized by the calling agent. The Python script itself (scripts/web_search.py) appears benign, correctly handling API keys via environment variables and communicating with the official Moonshot API (api.moonshot.cn) to perform its stated function.
Capability Assessment
Purpose & Capability
Name/description ask for a websearch capability and the skill requests Moonshot/Kimi API keys (KIMI_API_KEY or MOONSHOT_API_KEY). The script calls the Moonshot (api.moonshot.cn) chat/completions API and uses a Kimi model name, so requested credentials and endpoints are coherent with the stated purpose.
Instruction Scope
SKILL.md instructs running scripts/web_search.py which in turn calls the Moonshot API. The script does not perform local crawling itself; it relies on the remote model/tooling ($web_search) provided by Moonshot to obtain search results and URLs. This delegation is reasonable for a hosted websearch integration, but you should understand that user queries and the conversation context are transmitted to the Moonshot endpoint.
Install Mechanism
There is no install spec that downloads arbitrary code; the skill is instruction-only with a single included Python script. SKILL.md lists a pip dependency on openai which matches the script's imports. No high-risk external download URLs or extract steps are present.
Credentials
Only KIMI_API_KEY and MOONSHOT_API_KEY are required (primary is MOONSHOT_API_KEY). Those are expected for a Moonshot/Kimi integration and there are no unrelated secrets requested.
Persistence & Privilege
The skill is not forced-always, and does not request elevated persistence or modify other skills or system configs. Autonomous invocation is enabled (default) but not combined with additional concerning privileges.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install kimi-websearch - After installation, invoke the skill by name or use
/kimi-websearch - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.9
- Added new script: scripts/web_search.py; removed obsolete files for cleaner structure.
- Updated documentation with clearer API key setup, enhanced script usage instructions, and output requirements.
- Added metadata for integration (emoji, dependencies, required environment variables).
- Output must now include source URLs for transparency and reliability.
- Improved organization of news categories and formatting guidance for responses.
v1.0.8
- Replaced the main search script with a new file: scripts/kimi-web_search.py.
- Updated documentation to reflect the new script and simplified usage instructions.
- Removed details about deprecated script (scripts/web_search.py).
- Clarified environment variable requirements and cleaned up dependencies section.
- Streamlined output and process instructions for clarity.
v1.0.7
- Updated primary script: replaced scripts/kimi-web_search.py with scripts/web_search.py.
- Updated usage instructions: reflect the new script filename and clarified how to run the script.
- Added guidance for running with or without uv environment.
- Enhanced output requirements: now mandates providing source URLs and suggests example formatting for news-type inquiries.
- General documentation improvements for clarity and user experience.
v1.0.6
- Added embedded metadata for better environment and dependency management.
- Now documents primary API key as MOONSHOT_API_KEY and openai>=1.0.0 as a dependency.
- Updated usage paths to use `{baseDir}` for improved clarity.
- Removed obsolete file `_meta.json`.
v1.0.5
- Made minor documentation updates in SKILL.md for clarity and structure.
- Added a new _meta.json metadata file.
- Cleaned up metadata and requirements section for easier reading.
v1.0.4
- Added explicit pip dependency specification for the openai package (version >=1.0.0) in the skill metadata.
- Replaced previous install instructions under metadata with a standard dependencies list.
- No changes to instructions or script usage.
v1.0.3
- Added a new metadata section specifying emoji, required environment variables, and installation instructions.
- Declared MOONSHOT_API_KEY as the primary environment variable.
- Detailed the required Python package and installation method under metadata.
- No functional changes to the skill logic or code.
v1.0.2
- Added required environment variable documentation (`KIMI_API_KEY`, `MOONSHOT_API_KEY`) to SKILL.md.
- Listed `openai` as a dependency in SKILL.md.
- Updated script paths in usage instructions to support `{baseDir}`.
- Improved clarity in requirements and instructions for using API keys.
v1.0.0
- Initial release of kimi-websearch, a web search tool for retrieving information online.
- Supports queries via command-line using `kimi-web_search.py`.
- Requires setting either `KIMI_API_KEY` or `MOONSHOT_API_KEY` as an environment variable.
- Outputs concise, summarized answers based on search results, and not raw script output.
- Guidance provided for refining questions and handling incomplete or time-sensitive results.
Metadata
Frequently Asked Questions
What is Kimi Websearch?
联网搜索工具。Use When (1)用户需要通过联网获取信息; (2)如果你无法直接回答用户的问题，或者需要更多信息来回答用户的问题。 It is an AI Agent Skill for Claude Code / OpenClaw, with 591 downloads so far.
How do I install Kimi Websearch?
Run "/install kimi-websearch" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Kimi Websearch free?
Yes, Kimi Websearch is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Kimi Websearch support?
Kimi Websearch is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Kimi Websearch?
It is built and maintained by henryzhuhr (@henryzhuhr); the current version is v1.0.9.
More Skills