vss-generate-video-report

Generates structured video analysis and incident-range reports from VSS clips and analytics.

2|Updated Aug 20, 2026
One-click install
npx skills add https://github.com/atomicrajat/industry_safety_monitoring_system --skill vss-generate-video-report-atomicrajat
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: vss-generate-video-report
Source: https://github.com/atomicrajat/industry_safety_monitoring_system/tree/main/.claude/skills/vss-generate-video-report
Command: npx skills add https://github.com/atomicrajat/industry_safety_monitoring_system --skill vss-generate-video-report-atomicrajat

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) components.

What problem does it solve? Producing a consistent, structured report from NVIDIA VSS deployments requires choosing the right backend (VLM clip analysis vs. video-analytics incident queries), resolving clip URLs, and filling report templates correctly — this Skill automates that routing and rendering workflow. ## Core Features & Use Cases - Mode A — Per-clip VLM reports: Resolves a recorded clip URL via VST, calls an OpenAI-compatible VLM chat/completions endpoint (NIM Cosmos or RT-VLM), and renders a Video Analysis Report from a fixed template. - Mode B — Incident-range reports: Queries video_analytics__get_incidents through VA-MCP for a time range and sensor, tallies verdicts, and renders an Incident Range Report without calling the VLM. - Deployment-aware routing: Probes VST, VLM, and VA-MCP endpoints, hands off to deployment or storage skills when prerequisites fail, and rewrites clip URLs to browser-reachable public host/port values. - Use Case: An operator asks for a report on incidents on warehouse_safety_0001 between two timestamps; the Skill queries analytics, groups incidents by sensor, and returns a markdown report with verdict tallies and browser-playable evidence clip links. ## Quick Start Ask the agent to generate a report for a recorded sensor video or for incidents within a specific time range on a given sensor.

Frequently Asked Questions about vss-generate-video-report

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a video analysis report from a recorded VSS clip?

Use Mode A: resolve the clip URL through the VST storage API, then call the VLM chat/completions endpoint with a video_url content block. The result is rendered into the video-analysis-report.md template with a browser-playable clip link.

How do I report on incidents within a time range on a sensor?

Use Mode B: query video_analytics__get_incidents through the VA-MCP endpoint with ISO 8601 start and end times plus the sensor id. The Skill tallies confirmed, rejected, and unverified verdicts and renders the incident-range-report.md template without calling the VLM.

Which VLM backends does the report workflow support?

It supports NIM Cosmos and RT-VLM Cosmos, both exposing OpenAI-compatible chat/completions APIs. The backend is selected from the vss-agent container environment variables, with /v1/models probed before sending requests.

What happens when an incident query returns zero results?

The Skill returns a one-line statement naming the requested range and scope. It does not render the full template, invent incidents, seed test data, or fall back to Mode A VLM analysis.

When should I not use this report generation skill?

Avoid it for ad-hoc visual Q&A on clips, archive similarity search, read-only metrics lookups, deployment changes, or real-time alert rule management. Those requests route to dedicated ask, search, analytics, deploy, or alert skills instead.