Kits bundle skills, MCP servers, and project-specific instructions into a single activatable unit. Instead of configuring skills one by one, activate a kit to get a complete development setup tailored to your stack.
A kit combines three things:
| Component | Purpose |
|---|---|
| Skills | Claude Code skills that teach the agent how to perform specific workflows |
| MCP Servers | Tool servers that give Claude real-time access to your tools (dbt, Snowflake, etc.) |
| Instructions | Project conventions and best practices injected into every conversation |
| Kit | Description | Skills | MCP |
|---|---|---|---|
| dbt-snowflake | Complete dbt + Snowflake setup | 9 skills | dbt MCP server |
# Install the kit from this repo
altimate-code kit install AltimateAI/data-engineering-skills
# Activate a kit for your project
altimate-code kit activate dbt-snowflake
# Check what's currently active
altimate-code kit statusKits can detect when they're relevant to your project. For example, the dbt-snowflake kit looks for dbt_project.yml in your workspace and prompts you to activate it automatically.
kits/
└── my-kit/
└── KIT.yaml
name: my-kit
description: One-line description of what this kit sets up
version: 1.0.0
author: Your Name
tier: community
skills:
- source: "owner/repo"
select:
- skill-name-1
- skill-name-2
mcp:
tool-name:
type: stdio
command: ["uvx", "my-tool-mcp"]
env:
MY_VAR: "default-value"
env_keys: ["MY_VAR"]
description: "What this MCP server provides"
instructions: |
Project conventions injected into every conversation.
- Convention 1
- Convention 2
detect:
- files: ["config-file.yml"]
message: "Detected my-tool project — activate my-kit?"altimate-code kit validate my-kitSee CONTRIBUTING.md for submission guidelines.
| Field | Required | Description |
|---|---|---|
name |
Yes | Kit identifier (lowercase, hyphens) |
description |
Yes | One-line description |
version |
Yes | Semver version string |
author |
Yes | Author name or organization |
tier |
Yes | Trust tier: built-in, verified, or community |
Array of skill sources to install. Each entry has:
| Field | Description |
|---|---|
source |
GitHub repo in owner/repo format |
select |
List of skill names to install from that repo |
Map of MCP server configurations. Each entry has:
| Field | Description |
|---|---|
type |
Server transport type (typically stdio) |
command |
Command and arguments to start the server |
env |
Default environment variables |
env_keys |
Environment variable names the user may need to configure |
description |
What the MCP server provides |
Free-form text appended to Claude's context for every conversation in the project. Use this for team conventions, naming patterns, or workflow rules.
Array of auto-detection rules. Each entry has:
| Field | Description |
|---|---|
files |
Glob patterns — if any match, the kit is suggested |
message |
Prompt shown to the user when files are detected |
| Tier | Meaning | Who can publish |
|---|---|---|
built-in |
Ships with Altimate Code or this repo | Altimate AI |
verified |
Reviewed and approved by Altimate | Partner organizations |
community |
User-contributed, not reviewed | Anyone |
The tier is displayed when activating a kit so users can make informed trust decisions.
- Available skills in this repo
- CONTRIBUTING.md for contribution guidelines
- Altimate MCP Server Docs for MCP server details