What problem does it solve? Jetson diagnostics are scattered across tegrastats, jtop, procrank, nvpmodel, NvMap debugfs, and systemctl, each reporting a different slice of device state. This Skill replaces that guesswork with a single read-only snapshot so agents can answer identity, memory, GPU, thermal, power, and service questions from live data instead of assumptions. ## Core Features & Use Cases - All-in-one JSON snapshot: scripts/snapshot.sh emits SKU, variant, L4T version, memory, thermal zones, power mode, disk usage, top processes, GPU memory attribution, and candidate services in one structured payload. - Correct GPU memory source per driver stack: a capability probe picks nvidia-smi --query-compute-apps on the unified nvidia.ko stack (Thor) and falls back to /sys/kernel/debug/nvmap/iovmm/clients on the nvgpu stack (Orin), with gpu_source naming exactly which datum was used. - Quick human-readable summary: scripts/mem_summary.sh prints a one-line RAM/GPU/swap summary with an optional watch mode. - Use Case: A user says their Jetson feels slow and hot. The agent runs the snapshot, reports thermal zone temperatures alongside the top memory and GPU consumers, then hands off to an action skill like jetson-memory-audit or jetson-headless-mode without mutating anything itself. ## Quick Start Ask the agent to run the jetson-diagnostic snapshot script and summarize this Jetson's SKU, memory, thermal, and top processes.
