What problem does it solve? Finding the right pre-compiled model, branch, and sample for a specific D-Robotics RDK board (X5, S100/S100P/S600, X3, legacy) is error-prone: manifests vary by release, performance numbers are platform-specific, and running samples blindly can trigger unwanted system changes. This Skill grounds model discovery, download, and sample execution in the target repository's actual README, manifests, and code instead of guesswork. ## Core Features & Use Cases - Version-aware model discovery: Reads the target workspace's models.yaml manifest (via a read-only catalog script) to list available models, variants, and download conditions without inferring missing data. - Published benchmark lookup: Reports only recorded performance figures bound to their original platform and conditions, never copying numbers across boards or estimating missing values. - Guided sample execution: Reviews runtime README, entry points, and scripts before running, records cwd/argv/exit codes as scoped evidence, and requires user authorization for installs or system writes. - Toolchain handoff gating: Routes quantization or compilation needs to the appropriate X5 or S toolchain router only when available, otherwise guides installation through the pack installer. - Use Case: A user asks which YOLO model runs on their RDK X5 and at what frame rate. The Skill reads the target repo's manifest, selects a real sample with verified assets, reports only the published X5 benchmark, and provides the exact commands to run it. ## Quick Start Ask the agent which ready-made models are available for your RDK board and how to run one, pointing it at your local rdk_model_zoo workspace.
