Introduction to Skills
Agent Skills are skill packages that follow the agentskills.io open standard. Linkly AI Skills teach your AI assistant how to connect to and use Linkly AI’s tools, enabling AI to search, browse, and read your local documents. Why Do You Need Skills? After installing Linkly AI Desktop, you already have powerful document indexing and retrieval capabilities. However, AI assistants (such as Claude Code, Codex) don’t know how to use these tools. Skills act as an “instruction manual” that lets AI assistants automatically master the correct usage, including:- How to detect and connect to Linkly AI
- When to search, when to browse outlines, and when to read full text
- How to handle paginated reading of long documents
- How to safely handle document content
Installing Skills
- skills.sh (Recommended)
- Linkly AI Desktop
- Linkly CLI
- Claude Code
- Codex CLI
- Claude.ai Web
Use skills.sh to install to all supported AI assistants with one command:Or install to a specific AI assistant:
Using Skills
After installing Skills, you don’t need to learn any special commands. Simply ask questions in natural language through your AI assistant, and Skills will automatically recognize your intent and trigger the appropriate actions.Automatic Activation
The AI assistant will automatically determine whether to use Linkly AI’s tools based on keywords in your request. For example:Search
“Search my documents for content about API design”
Browse
“Show me the chapter structure of this report”
Read
“Read my project requirements document”
Combined
“Find my machine learning notes and summarize the core concepts”
How It Works
When the AI assistant receives your request, Linkly AI Skills automatically completes the following steps:1
Environment Detection
Independently detects both access paths — the CLI (
linkly command) and the MCP tools — and prefers the CLI when both are available.2
Search Documents
Generates appropriate search keywords based on your description and calls the
search tool to find relevant documents.3
Browse Outline
For long documents, first views the structure via the
outline tool to locate the sections that need to be read.4
Read Content
Uses the
read tool to retrieve the full content of the target sections, then summarizes, analyzes, or answers questions based on your needs.Using Skills for Integration
You can even directly tell the AI assistant to complete the Linkly AI integration. Simply send a prompt like the following:Please install the Agent Skill linkly-ai-skills and complete the LinklyAI tool integration...
- Independently detect whether the CLI and MCP access paths are available
- Pick a working mode based on what it finds (preferring the CLI when both are available)
- Run
linkly statusto verify the connection to the desktop application, falling back tolinkly doctorfor deeper diagnostics - If neither path is available, point you to the CLI installation guide
For security reasons, the Skill will not download or install the CLI on its own — it only prompts you to install it when needed. If you want an AI assistant to walk you through the whole setup from scratch, see the Setup Guide for Agents.
Security Notes
Linkly AI Skills security principles:- Clear scope — By default, only the indexed documents on your computer are accessed; your linked cloud libraries are accessed only when connecting remotely through
mcp.linkly.ai - Content isolation — Document content is only read as data, never executed as instructions
- Original files untouched — Skills only search, browse, and read your documents, and never modify them; the only write operation is creating or editing Linkly AI notes when you ask for it
- Credentials stay out of the conversation — Don’t paste your API key (starting with
lkai_) or LAN access token to an AI assistant; conversations are sent to the model provider and kept in the chat history. For remote access, let the AI assistant runlinkly auth loginand approve in your browser. If a key has already appeared in a conversation, revoke it and create a new one in the dashboard

