Creating Your First Skill

This guide walks you through creating your first Anava skill. Skills define what the AI looks for and how it responds. Once you create a skill and link it to a profile in a group, you'll have a working Detection.

What You'll Create

We'll create a simple but effective Intruder Detection skill that:

• Detects people in the camera view
• Determines if they appear authorized
• Triggers an ONVIF event for your VMS
• Optionally plays a voice warning

Prerequisites

• Access to Anava dashboard
• At least one group created
• Camera(s) assigned to the group

Step 1: Navigate to Skills

1. Open the Anava dashboard
2. Click Detections in the sidebar (or use AI Configuration > Detections)
3. Click Skills in the breadcrumb navigation
4. Click + Create Skill

Step 2: Basic Information

Fill in the basic details:

Field	Value	Notes
Name	"Intruder Detection"	Descriptive, unique within group
Description	"Detects unauthorized persons in secured area"	Helps identify skill purpose
Category	Security	Organizational tag

Step 3: Configure Analysis Prompts

Prompts guide the AI's analysis. Think of them as instructions to a security analyst.

System Prompt

The system prompt sets the AI's role and context. Enter:

You are a security camera analyst monitoring a secured facility.
Your job is to identify people who may not be authorized to be
in this area and assess any potential security concerns.

Be concise and action-oriented in your responses.

User Prompt

The user prompt gives specific analysis instructions. Enter:

Analyze this camera frame and:

1. Identify any people visible in the scene
2. Note if they appear to be employees (uniforms, badges, familiar behavior)
   or potential intruders
3. Describe their location and activity
4. Assess the threat level: none, low, medium, or high

Provide a brief summary suitable for a security operator.

Effective Prompts

Good prompts are:

• Specific about what to look for
• Clear about the output format
• Appropriate for the scene context

Step 4: Add Objects to Detect

Objects become ONVIF events that your VMS can receive.

Add "Person" Object

1. Click + Add Object
2. Configure:

Setting	Value	Purpose
Name	Person	Object identifier
Enabled	Yes	Active detection
Stateful	Yes	Sends True/False to VMS
Draw Bounding Box	Optional	Visual highlighting

Why Stateful?

Stateful objects emit:

• True when detected
• False when no longer detected

This allows VMS rules like "Record while Person is True".

Step 5: Add Questions (Optional)

Questions generate structured data beyond object detection.

Add Authorization Question

1. Click + Add Question
2. Configure:

Setting	Value
Name	is_authorized
Question Text	"Does this person appear to be authorized (badge, uniform, employee behavior)?"
Type	Boolean
Enabled	Yes

Add Description Question

1. Click + Add Question
2. Configure:

Setting	Value
Name	description
Question Text	"Briefly describe what you see"
Type	String
Enabled	Yes

Step 6: Configure TTS (Optional)

If you want the camera to speak when intruders are detected:

Enable TTS on Question

1. Select the "is_authorized" question
2. Enable TTS on Answer
3. Configure:

Setting	Value
Trigger When	Answer equals False
Voice	Kore (or your preference)
Style	[authoritative, warning]
Message Guidance	"Warn that this is a restricted area and security has been notified"

Step 7: Save and Test

1. Click Create Skill
2. Verify skill appears in skill list
3. Note the skill ID for profile creation

Test with Sample Image

Before connecting to profiles:

1. Click skill name to open details
2. Click Test Skill
3. Upload or capture a test image
4. Review analysis results

Step 8: Create a Profile (Next Step)

Your skill is ready! Now you need to create a profile that:

• Links a trigger (motion, I/O, etc.) to this skill
• Defines when analysis runs
• When assigned to a group, becomes a Detection

Understanding Your Skill

Here's what happens when this skill runs:

Skill Configuration Summary

Name: Intruder Detection
Category: Security

Analysis:
  System Prompt: Security analyst role
  User Prompt: Identify persons, assess authorization

Objects:
  - Person (stateful, triggers ONVIF)

Questions:
  - is_authorized (bool, triggers TTS if false)
  - description (string, for logging)

TTS:
  Voice: Kore
  Style: [authoritative, warning]
  Trigger: is_authorized = false

Common First Skill Issues

No Detections

• Verify skill is linked to active profile
• Check camera is online
• Ensure trigger type matches your test

Wrong Objects Detected

• Review and refine prompts
• Add more specific context
• Consider pre-filter for relevance

TTS Not Playing

• Verify TTS is enabled on question
• Check trigger condition matches answer
• Ensure camera has speaker

Next Steps

1. Create a Profile - Link this skill to a trigger
2. Best Practices - Learn detection design patterns
3. Tuning - Reduce false positives

Related Topics

• Detections Overview - How skills power detections
• Skills Reference - Complete skills documentation
• Settings Reference - All skill settings