Edge Impulse Agent Skills

This repository contains the public catalog of Edge Impulse Agent Skills.

Agent Skills are reusable building blocks that help agents interact with Edge Impulse data, projects, deployments, and workflows.

This repository is intended for developers and users who want to discover, install, and use skills. Every skill follows the Agent Skills specification.

Stable skills

Skill	Purpose
api	Edge Impulse APIs, CLIs, & SDKs
build-arduino-uno-q-app-lab	Arduino UNO Q and App Lab apps, bricks, Edge Impulse model integration, and Flask UIs
firmware-arduino	Arduino sketches, Edge Impulse library integration
build-custom-deployment-blocks	Edge Impulse custom deployment blocks

Experimental skills

Skill	Purpose
build-arduino-router-rpc	Custom MessagePack RPC clients for the Arduino UNO Q MPU/MCU router
build-custom-learning-blocks	Edge Impulse custom learning blocks (custom ML training containers)
firmware-cpp	Desktop, Linux, and generic MCU C++ apps using exported Edge Impulse libraries
firmware-raspberry-pi-python	Raspberry Pi and Linux Python inference using EIM models
firmware-stm32	STM32CubeIDE and FreeRTOS C++ inference integration
firmware-zephyr	Zephyr and nRF Connect SDK module integration

Installing and removing skills

The recommended way to install and manage skills is via the skills npm package developed by Vercel.

Below are some of the common commands for the skills CLI tool.

List all available skills:

npx skills add edgeimpulse/agent-tools --list

Install a skill (interactively):

npx skills add edgeimpulse/agent-tools

Install a specific skill:

npx skills add edgeimpulse/agent-tools --skill <skill-name>

Remove a specific skill:

npx skills remove <skill-name>

Catalog lifecycle

The lifecycle directories are repository conventions rather than Agent Skills frontmatter fields:

• Stable skills live under skills/
  • Backwards compatibility is expected.
• Experimental skills live under skills/.experimental/
  • They may change or be removed.
  • They are suitable for early adopters and feedback.
• Deprecated skills live under skills/.deprecated/.

Skill versions

Each SKILL.md stores a semantic version in metadata.version. Stable skills start at 1.0.0; experimental skills start at 0.1.0.

After a push to main, the version workflow increments the patch component once for every skill with changed files and opens a pull request containing those bumps. Changes outside a skill directory do not affect skill versions. Merging a generated version-bump pull request does not trigger another bump.

Contributing

To report a bug:

• File an issue in this repository

If you would like to propose a new skill or improvements to an existing one:

• Open a pull request against this repository
• Follow the guidelines in CONTRIBUTING.md.