Title: Wave Skills
Locale: en
URL: https://sensorswave.com/en/docs/ai/wave-skills/
Description: Wave Skills for Cursor, Claude Code, and Codex, focused on tracking design and analytics workflows.
Wave Skills is a set of Sensors Wave-oriented AI Agent Skills built on top of [MCP Setup](mcp.mdx). It does not replace MCP. Instead, it turns the underlying MCP tools into more stable and reusable workflows.

The current Skills focus on two main task groups:

- **wave-tracking**: tracking design, event naming, server-side vs client-side decisions, and the Pipeline / Tracking Plan / SDK init / quality check / Dashboard loop
- **wave-analytics**: event analysis, funnel analysis, retention analysis, user lists, user sequences, and custom SQL

The root-level `SKILL.md` acts as a router and dispatches requests to the right sub-skill based on user intent.

## When to use it

Wave Skills is useful if you regularly do the following in an AI Agent tool:

- have an AI Agent help design a tracking plan and produce implementation-ready event tables
- have an AI Agent operate on project assets such as Pipelines, Tracking Plans, and Dashboards
- have an AI Agent follow a repeatable workflow for event, funnel, retention, or SQL analysis
- want the AI Agent to answer in a way that is closer to Sensors Wave conventions, tool sequencing, and deliverable formats

## How it relates to MCP

- **MCP** provides the low-level project, tracking, analytics, and dashboard tools
- **Wave Skills** defines when to call which tools, what to confirm first, and how to structure the output

In practice:

- MCP answers "what can the AI Agent call?"
- Wave Skills answers "how should the AI Agent work?"

## Skills structure

Based on the current `agent-skills` repository, the main pieces are:

- root `SKILL.md`: umbrella entry that routes requests to tracking or analytics sub-skills
- `skills/wave-tracking/`: tracking conventions, interactive confirmation flow, SDK integration, and validation guidance
- `skills/wave-analytics/`: analysis model selection, MCP query construction, and result interpretation guidance
- `docs/README.install.md`: installation guide
- `install.sh`: installer for Cursor, Claude Code, Codex, or a project-local setup

## Installation

A typical installation flow looks like this:

```bash
git clone https://github.com/sensorswave/agent-skills.git
cd agent-skills
./install.sh --all
```

You can also install to a single platform or the current project only:

```bash
./install.sh --cursor
./install.sh --claude
./install.sh --codex
./install.sh --project
```

If you prefer manual setup, you can also symlink each Skill directory into the relevant AI Agent tool's `skills` directory.

## Recommended setup order

Use this order:

1. Complete [MCP Setup](mcp.mdx)
2. Install Wave Skills
3. Ask the AI Agent tool for a business task such as "design a signup-to-paid tracking plan" or "analyze 7-day retention for users acquired in the last 30 days"

That way, the AI Agent not only has access to Wave tools and project data, but can also follow a workflow that is closer to how teams actually operate.