Getting Started
Installation
Option A — clone and go. Working directly in the repo,
all skills are available immediately — they're auto-discovered as project skills under .claude/skills/.
Option B — install as a Claude Code plugin. To use these skills from any project, without cloning the repo as your working directory, add the marketplace and install the plugin:
/plugin marketplace add wilhelm-tiger/joyai-vl-interaction-skills
/plugin install joyai-vl-interaction@joyai-vl-interaction-skills
Either way, to provision the AWS EC2 instance and launch the model, tell your agent:
Create a new machine
This invokes the inference-machine-provisioning skill. The minimal service stack (WebInfer main model, summarizer
backend, and HTTPS WebUI) launches successfully and becomes accessible at the HTTPS WebUI URL:
https://<ec2-public-ip>:8099.
Requires AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables. If they aren't set, the skill will
either ask you to set them or automatically create a scoped IAM principal and print the credentials for future
sessions.
Since the server uses a self-signed SSL certificate, your browser will show a security warning. You can safely click Advanced → Proceed to site.
Want a real certificate on a real domain instead? See Networking.
What's Next
Once the machine is up, several optional skills extend it further:
- Networking — put the WebUI behind a real domain name with a trusted HTTPS certificate.
- RTSP Streaming — feed the model a looping local video file instead of a live webcam.
- Delegation & Screenshot Enhancement — have the model watch for a specific event, delegate to a background code-execution agent, and produce a polished screenshot of the moment.