What problem does it solve? Validating RDK Model Zoo samples across boards, runtimes, and model variants often produces vague or fabricated pass claims. This Skill enforces scoped, reproducible verification with explicit evidence, so smoke, accuracy, performance, and regression results are traceable to real commands, commits, and files. ## Core Features & Use Cases - Scoped Validation Planning: Builds a check matrix bound to the target commit, platform (X5, S100/S100P/S600, X3, legacy), model variant, and runtime, marking blocked checks as not-run instead of faking passes. - Structured Evidence Reports: Records argv, cwd, exit codes, timestamps, and SHA-256 hashed evidence files in a JSON receipt conforming to a bundled schema. - Read-Only Receipt Validator: A Python script checks report structure and evidence hashes without executing any commands or certifying hardware results. - Use Case: You fixed a preprocessing bug in a detect sample but have no board. The Skill runs available host/static checks, marks board-level rows not-run with reasons, and produces a verification receipt for review. ## Quick Start Ask the agent to validate the detect sample in your local rdk_model_zoo checkout and produce a verification report with evidence.
