Queries incidents, alerts, sensor IDs, places, and aggregated metrics stored in Elasticsearch via the VA-MCP JSON-RPC endpoint (default port 9901). It provides a two-step CLI pattern to initialize an MCP session and then call tools such as get_incidents, get_sensor_ids, get_places, fov_histogram, and analysis helpers to return recent incidents or aggregated statistics.
Use this skill when you need authoritative, recorded-event data from a video analytics stack: incident reports, alert history, object counts (people, vehicles), occupancy, average speeds, or sensor/place listings. It is intended for operational queries where the agent should run the provided curl commands against a reachable VA-MCP endpoint and relay actual results rather than guessing.
Likely compatible with agent tooling that can run shell commands and HTTP requests (CLI-enabled agents such as Copilot/Codex/Gemini CLI or Claude Code).
This skill has not been reviewed by our automated audit pipeline yet.
MoE Expert-Parallel Overlap (Megatron-Bridge)
Guidance and configs to enable expert-parallel communication overlap in Megatron-Bridge for MoE models — use to hide dispatch/combine latency and improve throug
Megatron CI/CD Guide
Reference guide to Megatron-LM CI/CD: pipeline structure, PR scope labels, triggering internal CI, and steps for investigating CI failures.
