- Home
- Skills
- Workflow Automation
- Video Analytics Alerts an Operator Can Act On
Video Analytics Alerts an Operator Can Act On
Cut false alarms to under one per camera per night, measure precision and recall on your own site, and hand verified alarms over.
$39
Works with the AI Agents you already use
Video Analytics Alerts an Operator Can Act On
Example session with this skill installed
I have a camera on a fence line that triggers 50 times a night because of tree shadows and passing cars. How do I fix this and set up a proper operator alert?
- Read your context and instructions
- Compiled the video analytics alerts
I have analyzed the fence line false alarm issue. To reduce these triggers and create a verified alert, we must move from simple motion detection to a defined event rule and a physics-first tuning approach.
Video Analytics Alert Configuration
| Component | Specification | Rationale |
|---|---|---|
| Event Rule | Person + Zone A (Fence) + Inward + >3s + 20:00-06:00 | Eliminates daytime traffic and transient motion. |
| Scene Fix | Mask top 20% (Road) and adjust IR axis | Removes headlights and foliage shadows from the algorithm. |
| Tuning Step 1 | Minimum Object Size (10% Frame Height) | Filters out small animals or wind-blown debris. |
| Tuning Step 2 | Duration Threshold (3000ms) | Ensures only persistent movement triggers an alert. |
| Evidence Package | 5s Pre-event + Detection Frame + Rule ID | Allows operator verification in under 4 seconds. |
| Escalation | Human Verify -> Site Manager -> Patrol | Prevents unverified calls to emergency services. |
Performance Targets
- Target Precision: >90% (Fewer than 1 false alarm per camera per night).
- Target Recall: >98% (Must catch all human-sized targets crossing the zone).
Next steps
- Apply the zone mask to exclude the road and high-foliage areas identified in the footage.
- Record 24 hours of video with these settings to calculate the initial false-alarm-per-night metric.
- Configure the VMS to attach the 5-second pre-event clip to every outbound notification.
Connects securely to your tools. The creator never sees your data.
About this skill
Under one false alarm per camera per night — measured, not claimed.
Turn a camera fleet into alerts an operator can actually act on - intrusion, line crossing, loitering, object left behind, vehicle access - with false-alarm reduction, a verified-alarm workflow, escalation to a monitoring centre, and measured precision and recall instead of vendor claims. Use when a surveillance system raises too many alerts, misses real events, or must hand verified alarms to a remote centre or a patrol.
An analytics system is not judged on what it detects but on what an operator does with it at 3 a.m. Thirty alerts a night and the operator stops looking - that system is worse than none, because everyone believes the site is watched. So the skill writes every rule as object + zone + direction + duration + schedule, fixes the scene before the algorithm (headlights, rain under IR, spiders on the lens, vegetation), and tunes in a fixed order where sensitivity comes last.
Includes an EXPERT.md reference: the seven-day measurement protocol on your own labelled footage, field targets for recall, precision and false alarms per camera per night, the five rule parameters with starting values, the nuisance table by frequency, the alert-chain latency budget, the escalation table to test at night before go-live, and failure diagnosis.
Acceptable use: lawful, declared and signposted protection of premises, goods and people's safety. The skill refuses facial recognition, biometric identification or categorisation, emotion recognition, tracking a named individual across cameras, profiling by appearance or origin, covert deployment, and employee monitoring beyond what is declared and proportionate.
It will not identify a person, will not promise a detection rate it has not measured on the site, and will not replace the operator: analytics raise candidates, humans decide.
Does not support facial recognition or biometric ID. Requires historical site footage to calculate precision and recall metrics. Scope is limited to event definition and tuning logic.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 27 days ago
Permissions
No special permissions declared or detected