Managing Skills
All skill management happens in the Skills panel (puzzle piece icon → Installed section).
Starting and Stopping
Each installed skill card has lifecycle buttons on the right side:
| Button | Action |
|---|---|
| ▶ Start | Launch the skill process and load its ML model |
| ■ Stop | Gracefully shut down the skill |
| ↻ Restart | Stop and immediately re-start (useful after config changes) |
When you click Start, the console panel opens automatically to show the skill's startup output — model loading progress, GPU detection, and readiness messages.
A green pulsing dot means the skill is running and processing frames.
Configuration
Click the gear icon on a skill card to expand its configuration panel. Configuration options vary by skill, but common settings include:
→ Confidence threshold — minimum detection confidence (0–1) → Processing interval — how often to analyze frames → Model variant — choose between accuracy and speed → Custom parameters — skill-specific settings
After changing a setting, click Save and then Restart the skill for changes to take effect.
Not all skills expose configuration options. If a skill has no configurable parameters, you'll see "No configuration available for this skill."
Viewing Source Code
Click the code icon (angle brackets) on any installed skill card to open the Source Browser. This shows:
→ The skill's file structure → Read-only view of every file in the skill package → Python scripts, configuration files, model definitions
This is useful for understanding what a skill does, debugging issues, or learning how to create your own skills.
Console Monitoring
Click the terminal icon in the Skills panel header to toggle the console. The console shows:
→ Live output from running skills — stdout in white, stderr in red → Tab bar — switch between multiple running skills, each with its own tab → Status dots — green (running), red (error), gray (stopped) → Input line — send text commands directly to a running skill's stdin → Auto-scroll — follows output automatically; scroll up to pause, click the arrow to resume
The console can also display inline images — some skills (like VLM tests) embed images directly in their output.
Uninstalling
To remove a skill:
- Click the trash icon on the skill card
- A confirmation prompt appears: "Remove? Yes / No"
- Click Yes — the skill is stopped (if running) and removed completely
Uninstallation removes the skill code and its environment. Downloaded models may remain cached to avoid re-downloading if you reinstall.
Command Approval Rules
Click the shield icon in the Skills panel header to configure approval rules. These control which system commands skills are allowed to run:
→ Allow — specific commands run without prompting → Deny — specific commands are blocked entirely → Ask — you're prompted to approve or deny each time
This ensures skills can't run anything dangerous on your system without your explicit permission.