← 返回 Skills 市场

Kagi

Name: Kagi
Author: michaelasper

作者 michaelasper · GitHub ↗ · v0.1.0

cross-platform ⚠ suspicious

1480

总下载

当前安装

版本数

在 OpenClaw 中安装

/install kagi

功能描述

Perform web searches and generate summarized answers with citations using the Kagi API and FastGPT for higher-quality results or rate-limit fallback.

使用说明 (SKILL.md)

Kagi (API)

Use the bundled Python scripts to call Kagi’s API from the OpenClaw host.

Quick start

Create a token in https://kagi.com/settings/api
Export it for your shell/session:

export KAGI_API_TOKEN='…'

Run a search:

python3 scripts/kagi_search.py "haaps glass" --limit 10 --json

Or ask FastGPT (LLM + web search):

python3 scripts/kagi_fastgpt.py "Summarize the latest Haaps glass mentions" --json

Tasks

1) Web search (Kagi Search API)

Use when you need a normal ranked list of results (URLs/titles/snippets).

Command:

python3 scripts/kagi_search.py "\x3Cquery>" [--limit N] [--json]

Notes:

Defaults to printing a readable digest; use --json for raw API output.
The script automatically sets Authorization: Bot \x3Ctoken>.

2) Answer/summarize with citations (FastGPT)

Use when you want a short answer grounded in web results, including reference URLs.

Command:

python3 scripts/kagi_fastgpt.py "\x3Cquestion>" [--cache true|false] [--json]

3) Using Kagi as a drop-in for web_search

If Brave Search is rate-limited (429) or you want better results:

Use scripts/kagi_search.py to fetch results
Then use the main agent model to synthesize / summarize based on the returned URLs/snippets

Files

API reference snippets: references/kagi-api.md
Python client + CLIs: scripts/kagi_client.py, scripts/kagi_search.py, scripts/kagi_fastgpt.py

安全使用建议

This skill appears to be a legitimate Kagi API client, but the package metadata fails to declare the required API token. Before installing: 1) Treat the KAGI_API_TOKEN as a secret — the scripts will send it to https://kagi.com/api/v0, which is expected for this skill. 2) Prefer the author update the skill metadata to list KAGI_API_TOKEN (and mark it as the primary credential) so automated permission checks are accurate. 3) Review the included scripts yourself (they are short and readable) to confirm there are no extra endpoints or unexpected behavior. 4) Use a token with limited scope if possible and rotate it if you test the skill on sensitive systems. If you cannot or do not want to provide an API token, do not install/use the skill.

功能分析

Type: OpenClaw Skill Name: kagi Version: 0.1.0 The skill bundle is benign. It provides Python scripts to interact with the Kagi API for search and FastGPT functionality. The code correctly retrieves the Kagi API token from environment variables (`KAGI_API_TOKEN` or `KAGI_API_KEY`) and makes network requests exclusively to the legitimate Kagi API endpoint (`https://kagi.com/api/v0`). There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts in `SKILL.md` or other files. All actions are aligned with the stated purpose of using the Kagi API.

能力评估

⚠ Purpose & Capability

The included Python scripts and SKILL.md clearly implement a Kagi Search + FastGPT client (calling https://kagi.com/api/v0). That matches the implied purpose, but the skill metadata declares no required environment variables or primary credential even though the code requires a KAGI_API_TOKEN (and also accepts KAGI_API_KEY as a fallback). The missing declaration is an incoherence between purpose and declared requirements.

ℹ Instruction Scope

SKILL.md instructs the user to export KAGI_API_TOKEN and run the provided scripts; the runtime instructions stay within the advertised scope (making HTTP calls to the Kagi API and printing/summarizing results). There is no instruction to read unrelated files or contact endpoints other than kagi.com. Note: the instructions assume an API token even though metadata doesn't list it.

✓ Install Mechanism

This is an instruction-only skill with small local Python scripts and no install spec or external downloads. No archive extraction or remote installers are used, which minimizes install risk.

⚠ Credentials

Although the only secret the code needs is the KAGI_API_TOKEN (reasonable for this purpose), the skill metadata lists zero required env vars and no primary credential. The code also accepts an alternate env name (KAGI_API_KEY). The omission in metadata is a discrepancy that could confuse users and permissioning systems; otherwise the requested credential is proportional to the skill's functionality and is only sent to Kagi's API.

✓ Persistence & Privilege

Skill flags are normal (not always:true). The skill does not request persistent system privileges, modify other skills, or write installers; it simply contains local scripts invoked at runtime.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install kagi
安装完成后，直接呼叫该 Skill 的名称或使用 /kagi 触发
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

v0.1.0

Kagi skill v0.1.0 - Initial release: Integrates Kagi Search API and FastGPT for enhanced web research. - Provides Python scripts for both ranked search results and LLM-powered web answers with citations. - Supports programmatic web search and summarization via command-line interface. - Includes setup instructions for API token usage and example commands. - Suggested as an alternative when Brave Search is rate-limited or higher-quality results are needed.

元数据

Slug kagi

版本 0.1.0

许可证 —

累计安装 5

当前安装数 5

历史版本数 1

常见问题

Kagi 是什么？

Perform web searches and generate summarized answers with citations using the Kagi API and FastGPT for higher-quality results or rate-limit fallback. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 1480 次。

如何安装 Kagi？

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

Kagi 是免费的吗？

是的，Kagi 完全免费（开源免费），可自由下载、安装和使用。

Kagi 支持哪些平台？

Kagi 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 Kagi？

由 michaelasper（@michaelasper）开发并维护，当前版本 v0.1.0。