Skip to content
Get packs

Guide · Agent skill

How to use the New AI Agent Project Setup agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The new-project skill is a set of written instructions that an AI agent (such as Claude) follows to help you build a new AI agent that reports to the Olakai monitoring platform. It covers project setup, designing business KPIs before any code, creating a workflow and agent with the Olakai CLI, defining customData fields, and integrating the Olakai SDK. This guide explains what the skill does, how to install it, how to trigger it, and what you need before you start.

New AI Agent Project Setupandrewyng/context-hub · MIT

Quick start

  1. 1Install the Olakai CLI. Run npm install -g olakai-cli, then run olakai login to authenticate your account.
  2. 2Install the skill. Unzip the downloaded ZIP into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  3. 3Ask for a new Olakai agent. Describe the goal in plain language, such as setting up a new AI agent that reports to Olakai, so the agent starts the guided process.
  4. 4Design KPIs first. Choose agentic or assistive AI, list the business questions, and map them to fields and formulas before writing code.
  5. 5Create the workflow and agent. Run olakai workflows create, then olakai agents create with --with-api-key and keep the returned key secret.
  6. 6Register custom data configs. Create a config for each NUMBER or STRING field you need, then confirm them with olakai custom-data list.
  7. 7Integrate the SDK and validate. Add the Olakai SDK to your TypeScript project with the agent key stored as a secret, then check end-to-end that events and KPIs appear.

What is an agent skill and what does this one do?

An agent skill is a file of instructions that an AI agent reads and follows. This skill is a single file, SKILL.md, so the agent works through the guide step by step rather than running a fixed program.

When active, the skill walks you through creating a new AI agent that reports to Olakai. It asks you to define the business questions and KPIs first, then maps them to data fields, creates a workflow and agent using the Olakai CLI, registers custom data fields that feed KPI formulas, and integrates the SDK.

  • Designs business KPIs before any SDK code is written
  • Creates a workflow and an agent with the Olakai CLI
  • Defines customData fields that KPI formulas can use
  • Integrates the Olakai SDK and covers end-to-end validation

How do you install the new-project skill?

Download the ZIP from this page and unzip it. You can install the skill in one of two ways: put the unzipped folder in ~/.claude/skills/ to make it available to you, or put it in a project's .claude/skills/ folder to make it available only in that project.

You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into your project's .claude/skills/
  • Claude apps: Settings → Capabilities → Skills, then upload

Before you start: what the skill requires

Every step depends on the Olakai platform, so you need an Olakai account and its CLI. The skill assumes you have Node.js and npm for installing the CLI, and a TypeScript project for the SDK code.

  • Olakai CLI, installed with: npm install -g olakai-cli
  • An Olakai account, logged into the CLI with: olakai login
  • An Olakai agent API key, generated via the CLI when you create the agent
  • Node.js/npm and a TypeScript project for the SDK
npm install -g olakai-cli olakai login

How do you trigger the skill with a natural request?

Ask the agent in plain language to set up a new agent with Olakai. The skill's description covers building a new AI agent with Olakai monitoring from scratch, including project setup, SDK integration, KPI configuration and end-to-end validation, so requests that describe that goal are the right trigger.

Good requests name the goal and the Olakai platform, for example asking the agent to start a new AI agent project that reports to Olakai and to design its KPIs first.

  • "Set up a new AI agent project that reports to Olakai, starting with the KPIs."
  • "Help me create an Olakai workflow and agent for my document summarizer, then add the SDK."
  • "Walk me through defining customData fields for my agent's KPIs in Olakai."

What does the agent do step by step?

The guide asks you to decide the agent type first. Agentic AI (multi-step autonomous workflows such as research agents or data pipelines) is tracked as single events that aggregate internal LLM calls. Assistive AI (chatbots and copilots) is tracked as a separate event per interaction.

Next comes KPI design. Each agent should have 2-4 KPIs, and only fields you register as custom data configs can be used in formulas. The agent then creates the workflow, creates the agent, and registers custom data configs with the CLI, all before SDK code is written.

  • Choose agentic or assistive AI type
  • Write the business questions and map them to data fields and formulas
  • Create a workflow (every agent must belong to one, even if it is the only agent)
  • Create the agent with --with-api-key and keep the returned key secret
  • Register custom data configs with olakai custom-data create, then verify with olakai custom-data list
  • Define KPI definitions, starting from templates via olakai kpis templates
olakai workflows create --name "Your Workflow Name" --json olakai custom-data list --agent-id YOUR_AGENT_ID

Key rules to keep in mind

Olakai's customData is only for KPI variables and for tagging or filtering. Platform-tracked values such as session ID, user email, timestamps, token counts and model should not be duplicated as customData.

KPIs belong to one agent only. If you create another agent later, it needs its own KPI definitions. Formula names are case-insensitive, and NUMBER fields must be sent as numbers, not strings.

  • Only registered CustomDataConfig fields can be used in KPIs
  • NUMBER configs need numeric values such as 5, not "5"
  • KPIs cannot be shared or reused across agents
  • Built-in variables such as Prompt, Response, PII detected and CODE detected are always available

Safety notes: scripts and network access

This skill does not run scripts. It uses the network, because every step talks to the Olakai platform through its CLI and API. The generated agent API key (format sk_agent_...) is a secret. The guide has you embed it in your SDK code, so do not paste it into shared code, public repositories or chats.

  • Scripts: none included; the skill is written instructions only
  • Network: yes, the CLI and SDK communicate with Olakai
  • Keep sk_agent_ keys out of shared code and chats; store them as secrets

What people use it for

  • Starting a document summarizer agent with a workflow, agent and KPIs for throughput and success rate
  • Setting up governance and monitoring for a chatbot-style support assistant on Olakai
  • Building a checklist of KPIs and customData fields before writing SDK code for a multi-step research pipeline
  • Adding a second agent to an existing workflow, with its own KPI definitions

Frequently asked questions

Is the new-project skill free to use?

The source does not state pricing or licence terms beyond the MIT licence listed for the item. The skill itself is a set of instruction files, but using it requires an Olakai account and CLI access, so check Olakai's own terms for platform costs.

What do I need before installing this skill?

You need the Olakai CLI (npm install -g olakai-cli), an Olakai account logged in with olakai login, and Node.js/npm plus a TypeScript project for the SDK. Every step depends on Olakai's CLI and API keys.

Where do I put the skill files?

Unzip the downloaded ZIP into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.

Does this skill work without Olakai?

No. The skill is built around the Olakai platform, its CLI and API keys. If your project does not use Olakai, this skill is not the right tool.

Why do the KPI fields need to be registered before the SDK code?

Only fields registered as CustomDataConfigs become variables for KPI formulas. Unregistered customData fields are stored but cannot be used in KPIs, so the guide requires registration first.

Can I reuse the same KPIs across several agents?

No. Each KPI belongs to exactly one agent, so every new agent needs its own KPI definitions and custom data configs.

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 developers starting a new agent on Olakai who want a structured checklist for KPIs and customData before writing SDK code. It is a poor fit if you do not use Olakai, since every step depends on its CLI and API keys, and anyone who shares sk_agent_ keys in code or chats should not use it as written.

Download the skill