Guide · Agent skill
How to use the agent-squad TypeScript guide agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The agent-squad-typescript skill is a guidance file for AI coding assistants working on Node.js and TypeScript apps that use the agent-squad npm package. It gives the agent a map of the framework: how to set up an orchestrator that routes user requests to specialist agents, which agent type fits which job, and which pitfalls to avoid. After following this guide you will know how to install the skill, how to ask for it, and what it can and cannot do for you.
agent-squad TypeScript guide2FastLabs/agent-squad · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the agent-squad-typescript skill ZIP from this page.
- 2Install the skill. Unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Install the npm package. In your Node.js project, run npm install agent-squad, plus any optional peer dependencies you need.
- 4Ask for a specific task. Describe the task in plain language, naming agent-squad and what you want to build, such as a router between two agents.
- 5Check the source. Confirm exact method signatures and options in the library's typescript/src/ folder before you rely on generated code.
- 6Test the orchestrator. Run your routing with a few sample messages to confirm the classifier picks the agent you expect.
What the skill makes the agent do
Once installed, the skill tells the agent what agent-squad is for and what each part does. agent-squad is a multi-agent orchestration framework: an orchestrator (AgentSquad) receives each user message, a classifier picks the right agent, and that agent produces the reply. The skill covers the built-in agent types, including single assistants, GroundedAgent for answers that must match live data, ChainAgent for fixed pipelines, and SupervisorAgent for a lead agent that calls sub-agents as tools.
The skill is a map, not an API reference. It points the agent to the library's source code and documentation so it can check exact signatures before writing code. That is meant to keep generated code in line with the library's conventions.
Install the skill
You can install the skill in either of two ways. Download the ZIP file from the page, then unzip it into your personal skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it limited to one codebase.
Alternatively, upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal: unzip into ~/.claude/skills/
- Per project: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Install the agent-squad package in your project
The skill guides code for an app that already uses the agent-squad npm package. Add the package to your project with the command below. Optional peer dependencies are needed only for specific features.
- @modelcontextprotocol/sdk: needed for MCPToolProvider (MCP tools)
- @dakera-ai/dakera: needed for the DakeraRetriever
npm install agent-squadTrigger the skill with natural requests
You do not need a special command. Describe the task in plain language while working in a Node.js or TypeScript project that uses agent-squad. The skill's description says it applies when you are building or modifying an app that uses the agent-squad package for multi-agent orchestration, classifier routing, storage, retrievers, or tools.
Name the framework and the job. The more specific the request, the more useful the guidance.
- Set up an agent-squad orchestrator that routes between a billing agent and a tech support agent.
- Which agent type should I use for answers that must match live prices, GroundedAgent or ChainAgent?
- Add DynamoDB chat storage to my existing agent-squad orchestrator.
- Add an MCP tool to my BedrockLLMAgent using MCPToolProvider.
Understand the basic turn
The skill explains how a request flows through the framework. routeRequest is the single entry point. It classifies the input, sends it to the chosen agent, saves the exchange, and returns a response. The response is either a plain string or a streaming output, depending on the agent's settings.
Agent IDs matter. An agent's id is derived from its name (for example, "Tech Support" becomes "tech-support"), and that id is used for storage and classifier matching, so keep names stable across restarts.
- Single assistant: one agent, optionally with no routing at all.
- Several specialists: register agents with addAgent and let the classifier route each turn.
- Data-sensitive answers: GroundedAgent, where a gatherer calls tools and a separate presenter speaks only from curated results.
- Fixed pipeline: ChainAgent, where each output feeds the next agent.
- Lead with team: SupervisorAgent, where the lead calls sub-agents as tools.
Watch for common mistakes
The skill also flags pitfalls. Check the source before relying on any detail, because the guidance itself says signatures must be verified in the code. Some constraints the skill calls out: routeRequest does not throw, since errors come back as a non-streaming response with the error text in output; SupervisorAgent requires a BedrockLLMAgent or AnthropicAgent as its lead, and that lead must not have its own toolConfig; and when using MCPToolProvider you pass it as toolConfig.tool and omit useToolHandler.
- Verify signatures in typescript/src/ before shipping code.
- Keep agent names stable so agent ids do not change between restarts.
- Handle both streaming and non-streaming responses.
What people use it for
- Building a multi-agent customer support chatbot that routes billing and technical questions to different agents.
- Creating an assistant that answers price or balance questions from live tool results without the model inventing figures, using GroundedAgent.
- Adding chat history storage to an existing Node.js agent-squad app.
- Connecting MCP tools to a BedrockLLMAgent in an existing TypeScript project.
Frequently asked questions
What is the agent-squad TypeScript agent skill?
It is a guidance file for AI coding assistants that covers the agent-squad npm package for Node.js and TypeScript. It explains how to set up an orchestrator, which agent type to pick, and which mistakes to avoid. It points to the source code and docs rather than replacing them.
Do I need to know what agent skills are before using this?
No. Download the ZIP, unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps. After that, ask for agent-squad help in plain language.
What do I need to run it?
The skill itself is a file, and the code it helps with needs Node.js and npm, the agent-squad npm package, and AWS credentials for Bedrock-based agents or an Anthropic or OpenAI API key for those agent types. MCP and Dakera support need their optional packages.
Is the skill free, and what is its licence?
The item is listed under the Apache-2.0 licence. The source does not mention pricing, so check the listing for any cost details before you download.
Does the skill run scripts or use the network?
According to the listing, it runs no scripts and makes no network calls. It is guidance text that the agent reads while helping you write code.
Can I use this skill for a Python project?
No. The skill covers only the TypeScript version of agent-squad, so Python developers should look elsewhere for guidance.
Can I rely on it for exact API details?
No. The skill states it is not an API reference, so check method signatures and options in the library's source code before shipping anything.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits TypeScript developers who are building multi-agent chatbots or assistants with agent-squad on AWS Bedrock, Anthropic or OpenAI, and who want an AI coding assistant that avoids common mistakes. It is the wrong choice if your project does not use the agent-squad npm package, if you work in Python, or if you need authoritative API details without checking the library's source.
Download the skill