@aemonculaba/pi-search
Provides AI Agents with web search and content extraction capabilities, supporting OpenAI's native search and multi-mode fetching.
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: @aemonculaba/pi-search. It is used for: Provides AI Agents with web search and content extraction capabilities, supporting OpenAI's native search and multi-mode fetching. Full Skill content: https://321skill.com/skills/aemonculaba-pi-search/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
This Skill addresses the issue where AI Agents cannot directly access real-time web information during development and research. In practice, developers often need to query the latest technical documentation, obtain real-time data, or analyze competitor webpages, but traditional Agent tools often lack secure and efficient internet connectivity.
Usage is straightforward: simply install this extension in Pi, then invoke web_search via natural language commands for keyword searches, or use web_fetch to retrieve the content of a specified URL. You can then use natural language instructions to obtain a list of search results (including titles, URLs, and snippets) or the cleaned, Markdown-formatted page body processed by Readability.
It is particularly suitable for teams or individuals who frequently conduct online research and data analysis, such as full-stack developers, data analysts, and content creators. Especially for users already on the Pi platform looking to extend their AI assistant's external information-gathering capabilities, it can significantly improve the efficiency of information retrieval and content analysis.
It is recommended to use this tool consistently for tasks like market research, competitor analysis, or writing reports that require the latest information. Note that it primarily uses OpenAI's Codex subscription or API for searches, blocks access to unsafe targets like localhost and private IPs by default, and limits fetched response bodies to under 2MB to ensure performance and security.
Key Features
Unlike crawler tools that rely on public search engines like DuckDuckGo, it defaults to using OpenAI/Codex's native web_search API, avoiding the crawl restrictions and rate limits common with public search engines.
Limitations
The `web_search` feature relies by default on Pi's Codex OAuth subscription (model gpt-5.4-mini). To use an OpenAI API key (model gpt-4o), it must be manually enabled in the configuration. Additionally, the Playwright fetching mode requires a local Chromium installation.
FAQ
How do I enable the OpenAI API as a search fallback?
Set the environment variable PI_SEARCH_ENABLE_OPENAI_REGISTRY_FALLBACK=true or PI_SEARCH_ENABLE_OPENAI_ENV_FALLBACK=true, and ensure the corresponding API key is available.
Can web_fetch retrieve JavaScript-rendered pages?
Yes, it provides a fallback solution using Playwright + Readability, specifically designed to handle JS-heavy pages.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/aemonculaba-pi-search/raw/index.md to read the original Skill definition (Markdown format) for @aemonculaba/pi-search, and install it according to the instructions.
Raw Markdown URL for AI: /skills/aemonculaba-pi-search/raw/index.md