Release Notes: Skills#

Version 2026.2#

Release Date: September 9, 2026

The first release of the Skills repository introduces a catalog of user-facing agentic skills for deploying, operating, and extending Open Edge Platform components.

New:

  • DL Streamer: The dlstreamer-coding-agent skill guides creation of video analytics applications using Python, C, C++, or GStreamer.

  • Chat Question and Answer: The chatqna-docker-deploy and chatqna-helm-deploy skills guide deployment and operation of ChatQnA Core with Docker Compose and Kubernetes.

  • Video Search and Summarization: The vss-deploy, vss-deploy-helm, vss-search-index, and vss-summarize-video skills cover deployment, video indexing, semantic search, and video summarization workflows.

  • Multimodal Embedding Serving Microservice: The multimodal-embedding-serving-user skill covers service deployment and embedding text, images, and videos through its REST API or Python SDK.

  • Multimodal DataPrep Microservice: The multimodal-dataprep-user skill covers deployment, storage configuration, and ingestion and management of multimedia retrieval data.

  • Model Download: The model-download-user skill guides downloading and converting models from supported sources into deployment-ready formats.

  • DL Streamer Pipeline Server: The dlsps-user skill covers deploying the pipeline server and operating configured video analytics pipelines through its REST API.

  • Time Series Analytics Microservice: The time-series-analytics-user skill guides deployment and creation of streaming and batch analytics use cases, including alerts and model inference.

  • Physical AI Train: The physicalai-train-adding-a-policy, physicalai-train-benchmarking-a-policy, physicalai-train-exporting-and-validating, physicalai-train-training-a-policy, and physicalai-train-working-with-datasets skills cover the policy development lifecycle from datasets and training through benchmarking and export.

  • Physical AI Runtime: The physicalai-runtime-adding-a-camera-backend, physicalai-runtime-adding-a-robot-integration, physicalai-runtime-configuring-inference-pipeline, physicalai-runtime-loading-exported-policies, and physicalai-runtime-running-policy-on-robot skills cover camera and robot integration, inference configuration, policy loading, and hardware execution.

  • Geti: The geti-using-the-pipeline skill covers the project-to-deployment workflow, while the six getitune-* skills cover model discovery, dataset preparation, training, inference, export, and optimization.

  • SceneScape: The scenescape-setup skill guides end-to-end installation, configuration, calibration, and verification of a SceneScape deployment.

  • Metro AI Suite Prompt Library: The metro-ai-apps-builder skill translates a business objective into a complete Intel Edge AI application plan and delegates implementation to relevant skills.

  • Metro AI Suite Vision AI App Recipe: The metro-ai-apps-recipe skill deploys a computer-vision analytics stack with live video, dashboards, and alerts from a video source and model.

Improved:

  • Index Maintenance: The workflow now batches installs by (repo, ref) to avoid redundant clones and retries skill discovery by path for skills nested beyond the CLI’s default scan depth.

  • Stale Skill Detection: update_skills_index.py now detects stale skills from disk independently of skills-lock.json.

On this page

This Page