AI Agent Integration and Developer Experience#
The Edge Node Infrastructure Blueprint ships a set of agent skills that let you run platform workflows through natural language, using GitHub Copilot or Claude Code. Instead of manually running scripts and commands, describe the outcome you want and the agent takes care of the rest.
Available Skills#
Skill |
What it does |
|---|---|
|
Builds a host OS image using the Image Composer Tool or ISO based curation |
|
Packages a complete bootable USB artifact ( |
|
Validates a provisioned edge node over SSH — checks k3s pod health, binary paths, cloud-init state, network/proxy settings, and device readiness (GPU VFs, NPU) |
|
Applies a CPU and/or GPU power profile ( |
|
Updates Ubuntu package configuration and installs required packages on a provisioned system |
How to Use Skills#
Open GitHub Copilot Chat or Claude Code in the repository workspace and describe what you want in natural language. The agent matches your request to the appropriate skill, asks for any missing inputs, then runs the workflow. These skill have been verified by running them from the developer system and pointing to the provisioned target system whereever applicable.
Note that few skills like validate-platform-config and tune-platform-power run commands on the target system, hence it is expected that during configuration, a user has seeded a public key for passwordless access.
Example Prompts#
Build an OS image for my Intel Core Ultra system.
Create a bootable USB artifact using the ICT image at /path/to/image.raw.gz.
Validate the provisioned node at 192.168.1.10 — check pods, drivers, proxy settings, and GPU VFs.
Switch the node at 192.168.1.10 to the balanced power profile.
Add sysbench and stress-ng to ict-template only for a debug image variant.
What the Agent Does#
For every skill invocation the agent follows this fixed sequence:
Collects required inputs — prompts for anything not provided in the request
Runs preconditions — verifies the environment is ready before making any changes
Executes steps — runs the workflow, pausing for confirmation before any privileged or destructive action
Validates results — confirms the expected artifacts or system state are present
Reports outcome — returns artifact paths, validation status, and troubleshooting notes on failure
Note: The agent will never proceed past a failed precondition or skip the validation step.
Providing Required Inputs#
Each skill declares specific required inputs. If you omit them, the agent will ask before proceeding. Common inputs include:
Input |
Used by |
|---|---|
Image template path ( |
|
ICT image path ( |
|
Target node SSH address and credentials |
|
Power profile ( |
|
Target ( |
|