Skill: NVIDIA VSS Video Q&A
by NVIDIA
NVIDIA's official skill for asking a vision-language model fresh questions about recorded video through the Video Search and Summarization blueprint — only when the pixels actually need looking at.
NVIDIA's official skill for generating and validating runnable gst-launch DeepStream pipelines from a plain-language description of the video inference you want.
An official, NVIDIA-authored agent skill that builds NVIDIA DeepStream GStreamer pipelines. Hand-writing a gst-launch-1.0 line with nvstreammux, nvinfer, nvtracker and the right sink for your platform is famously unforgiving — one wrong property name and the pipeline dies with an error that says little about which element caused it. This skill turns that into a short interview plus a validator.
The agent collects requirements through an interactive questionnaire — up to seven questions, skipping any the request already answered — then assembles a pipeline using a self-contained BM25 retrieval backend with structural metadata boosting, searching over 270 or more verified reference pipelines. The retrieval engine is Python stdlib only, so there is nothing to pip install and no network call at generation time.
"detect and track on 4 rtsp streams and display on jetson" is enough to skip most questions. "build a pipeline" triggers all seven.
nvinfer, primary plus secondary, with a preprocessor, or Triton via nvinferserver.A bundled four-stage validator checks syntax, that every element exists, that each property is real for that element, and finally does a live parse. A separate linter keeps the underlying pipeline dataset clean, with an auto-repair mode.
Python 3.8+ (standard library only), the DeepStream SDK installed at /opt/nvidia/deepstream/deepstream/, and gst-launch-1.0 and gst-inspect-1.0 on PATH.
Part of NVIDIA's agent-skills catalogue, mirrored daily from NVIDIA's product repositories. Install everything with npx skills add nvidia/skills, or this skill alone with npx skills add nvidia/skills --skill deepstream-generate-pipeline --yes. Source is Apache-2.0, skill text is CC-BY-4.0.
Once installed, ask your agent something like:
"Give me a DeepStream pipeline that runs primary detection plus NvDCF tracking on four RTSP cameras and writes an MP4 per stream, on a Jetson Orin — then validate it before I run it."
The instructions your agent loads, as published by NVIDIA.
by NVIDIA
NVIDIA's official skill for asking a vision-language model fresh questions about recorded video through the Video Search and Summarization blueprint — only when the pixels actually need looking at.
by NVIDIA
Trains, distils, quantises, evaluates, exports and runs inference for RT-DETR real-time object detection in NVIDIA TAO, routing training through AutoML hyperparameter search by default.
by NVIDIA
Runs NVIDIA's NV-Reason-CXR-3B chest X-ray reasoning model through a documented wrapper, locally on GPU or via the public Hugging Face Space, and returns the model's full reasoning trace as engineering evidence.
by NVIDIA
Inspects the host's hardware, OS, CUDA driver and existing tooling, recommends one Holoscan SDK install method with a reason, then hands off to the matching install skill.