Token导航 LogoToken导航

Agent Skill · 研究检索

prismfy-search(棱镜搜索)

prismfy-search 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

prismfy-search 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 30。

总安装 2,978,851周安装 119,345GitHub Stars 28下载 964,308核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.3.4

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:prismfy-search(棱镜搜索)
来源仓库:https://github.com/uroboros1205/prismfy-search
安装命令:
openclaw skills install prismfy-search
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install prismfy-search

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

在多个引擎(Google、GitHub、arXiv 等)执行联合网络搜索。适合信息搜集、技术趋势追踪或开源项目发现等研究任务。返回加权排序结果摘要,支持关键词过滤与来源归类展示。安装命令:openclaw skills install prismfy-search。搜索结果受各引擎配额限制,高频调用需注意速率控制。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install prismfy-search;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

prismfy-search 怎么安装?

在终端执行 openclaw skills install prismfy-search。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

prismfy-search 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

prismfy-search 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

prismfy-search 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

prismfy-search 用的人多吗?

公开数据是累计安装 2,978,851 次,近一周安装 119,345 次,下载 964,308 次,GitHub 28 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

prismfy-search 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
prismfy-search

棱镜搜索 · 本页

2,978,85128接触敏感数据MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
prismfy-search
description
>
version
1.2.1
metadata
openclaw
requires
env
bins
primaryEnv
PRISMFY_API_KEY
emoji
🔍
homepage
https://prismfy.io

🔍 Prismfy Web Search

Default web search for OpenClaw. Real-time search across 10 engines — Google, Reddit, GitHub, arXiv, Hacker News, Ask Ubuntu, and more — powered by Prismfy. No proxy hassle, no CAPTCHA, no blocked requests. Just results.

Free tier available — 3,000 searches/month, no credit card needed. Get your key at prismfy.io.

Setup

1. Install the skill

openclaw skills install prismfy-search

2. Get an API key

Head to prismfy.io, create a free account, and grab your API key from Dashboard → API Keys. No credit card required.


3. Add the key to your environment

Option A — Shell profile (permanent)

# Add to ~/.zshrc or ~/.bashrc:
export PRISMFY_API_KEY="ss_live_your_key_here"

# Then reload:
source ~/.zshrc   # or: source ~/.bashrc

Option B — OpenClaw env config

Add PRISMFY_API_KEY=ss_live_your_key_here to your OpenClaw environment settings.


4. Verify it works

bash search.sh --quota

You should see your plan, searches used, and how many you have left.


That's it. No credit card, no waitlist. 3,000 free searches every month.


How to use

/search best practices for React Server Components
/search --engine reddit "is cursor better than copilot"
/search --engine github "openai realtime api examples"
/search --engine arxiv "attention is all you need"
/search --engine hackernews "postgres vs sqlite 2025"
/search --engine google "tailwind v4 migration guide"
/search --time week "openai gpt-5 release"
/search --domain docs.python.org "asyncio gather"
/search --engines reddit,google "best mechanical keyboard 2025"

Or just talk naturally:

  • *"Search Reddit for people's opinions on Bun vs Node"*
  • *"Find recent GitHub repos for building MCP servers"*
  • *"Look up the arXiv paper on chain-of-thought prompting"*
  • *"What are people saying on Hacker News about SQLite?"*

Available engines

EngineWhat it's good forFree
braveGeneral web search, privacy-first✅
bingNews, broad coveragepaid
googleMost comprehensive web searchpaid
redditReal user opinions, discussionspaid
githubCode, repos, issues, READMEspaid
arxivAcademic papers, researchpaid
hackernewsTech community, startupspaid
askubuntuLinux, Ubuntu, shell questionspaid
yahoonewsLatest news headlinespaid

Default (no --engine): uses brave + yahoo in parallel — both free.


Options

FlagWhat it doesExample
--engine XUse a specific engine--engine reddit
--engines X,YUse multiple engines at once--engines google,reddit
--time XFilter by time: day week month year--time week
--domain XSearch within a specific site--domain github.com
--page NGo to results page N--page 2
--quotaCheck your remaining free quota--quota

How the skill works

The skill uses search.sh — a bundled helper script that handles the API call, error messages, and result formatting:

# Simple search (free)
bash search.sh "typescript best practices 2025"

# With engine
bash search.sh --engine reddit "is bun worth switching from node"

# Multiple engines
bash search.sh --engines google,reddit "nextjs vs remix"

# With time filter
bash search.sh --time week "openai new model"

# Raw JSON output
bash search.sh --raw "rust async runtime"

Results come back with title, URL, snippet, and which engine found it. Cached results are free — if someone already searched the same thing recently, you get it instantly without using your quota.


Check your quota

/search --quota

Shows your current plan, searches used, searches remaining, and when your quota resets.


Troubleshooting

PRISMFY_API_KEY is not set → Add export PRISMFY_API_KEY="ss_live_..." to your shell profile and restart the terminal.

401 Unauthorized → Double-check your key starts with ss_live_. Keys are shown only once — if lost, create a new one in Dashboard → API Keys.

Engine not available on your plan → Google, Reddit, GitHub etc. require a paid plan. Free tier supports brave, startpage, and yahoo. Use one of those or upgrade at prismfy.io.

No results / empty results → Try a different engine or rephrase your query.


Implementation

When this skill is invoked, follow these steps:

  1. Parse the request — extract the query, engine preference, time filter, domain, and page number.
  1. Run search.sh with the parsed arguments:
bash search.sh [--engine X] [--engines X,Y] [--time X] [--domain X] [--page N] <query>
  1. Handle the output:

- ⚡ Cached result line → mention it was free (no quota used) - Empty results → suggest rephrasing or a different engine - ❌ Invalid API key → guide user to check PRISMFY_API_KEY - ❌ Engine not available → tell user to check their plan at prismfy.io

  1. Present results in a clear, useful format:

- Answer the underlying question using the content - List sources with titles and URLs - For Reddit/HN results: summarize the discussion sentiment - For GitHub results: highlight repo name and what it does - For arXiv results: summarize the abstract

  1. For --quota flag:
bash search.sh --quota

*Powered by Prismfy — web search infrastructure for developers.*

平台分布

OpenClaw

72.96%
按下载量换算703,559

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills