Generated banner for Jetson Diagnostic
Generated logo for Jetson Diagnostic

Jetson Diagnostic

NVIDIAOfficialCreated by Jetson TeamDistributed by NVIDIA

  • Skill
  • Tool integration
  • Cloud and DevOps
  • Apache-2.0
  • Medium risk
  • Has scripts

From the author

Read-only Jetson health snapshot for identity, memory, GPU, thermal, power, storage, services, and top processes.
— the description by Jetson Team

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

Jetson Diagnostic provides a unified, read-only interface for gathering telemetry from NVIDIA Jetson hardware. It aggregates data from various system utilities including tegrastats, NvMap, and system process monitors to provide a comprehensive snapshot of device identity, memory usage, GPU load, thermal state, power settings, and active services. It is designed to help agents identify system bottlenecks and resource consumption patterns on Jetson platforms.

What it does

  • Captures a consolidated JSON snapshot of Jetson system health.
  • Identifies Jetson SKU, generation, and L4T version.
  • Reports thermal, power, and disk usage metrics.
  • Provides per-process GPU memory attribution via NvMap or nvidia-smi.
  • Lists top processes and candidate services for resource management.
  • Offers a compact summary for quick memory and swap status checks.

When to use it

  • When investigating performance issues or unexpected system behavior on a Jetson device.
  • To establish a baseline measurement before running memory-intensive tasks or benchmarks.
  • When needing to identify which processes are consuming GPU or system memory.
  • To verify the current power mode and thermal status of the hardware.

Advantages

  • Unified output replaces the need to manually query multiple disparate system tools.
  • Automated detection of the correct GPU memory source based on the driver stack.
  • Provides actionable data for further troubleshooting with specialized Jetson tools.
  • Read-only design ensures no unintended changes are made to system configuration.

Usage examples

“What is the current health status of this Jetson device?”
The agent executes the snapshot script and summarizes the SKU, thermal readings, memory usage, and top processes.
“Why is my Jetson running slowly?”
The agent runs the diagnostic snapshot to report thermal levels and identify processes with high CPU or NvMap memory consumption.

Install

  1. Download the skill as a ZIP:

    https://mp-staging.autogineer.com/agent-skills/nvidia/jetson-diagnostic/download
  2. Unzip it into the cross-client skills folder of your project:

    mkdir -p .agents/skills && unzip jetson-diagnostic.zip -d .agents/skills

Documentation by Jetson Team

Jetson Diagnostic

A unified, agent-friendly view of a running Jetson device. Replaces the need to remember which of tegrastats, jtop, procrank, /sys/kernel/debug/nvmap, nvpmodel, free, swapon, df, and systemctl list-units produces which slice of the truth.

Purpose

Capture a read-only health snapshot from the Jetson host so agents can answer device identity, memory, GPU, thermal, power, storage, and service-state questions using live data instead of guesses.

When to use

Activate when the user asks:

  • "What is this Jetson? What SKU? How much memory?"

  • "What's running on this Jetson right now?"

  • "Why is my Jetson slow / hot / out of memory?"

  • "Give me a snapshot of GPU / CPU / power usage."

  • "What does my tegrastats output mean?"

  • "Which services are running that I could turn off?"

  • The user has installed jetson-memory-audit, jetson-headless-mode, jetson-inference-mem-tune, jetson-llm-benchmark, jetson-llm-serve, or jetson-package and needs a baseline measurement before running them.

Do not use this skill to change power modes, drop caches, stop services, install packages, serve models, or tune inference flags. Report the observed state, then hand off to the action-oriented skill.

Prerequisites

  • Run on the Jetson host, or in a sandbox/container with host-visible Jetson system paths and process data.

Available Scripts

ScriptPurposeArguments
scripts/snapshot.shEmits the all-in-one JSON snapshot for identity, memory, GPU, thermal, power, disk, top processes, and candidate services.--human, --tegra-secs N, --top-procs N.
scripts/mem_summary.shEmits a compact human-readable RAM/GPU/swap summary.--short, --watch, --interval N.
scripts/detect_jetson.shExports or prints canonical Jetson SKU/generation/product-line fields for this repo.No arguments.

If your agent runtime supports run_script, use it to run scripts/snapshot.sh or scripts/mem_summary.sh and summarize the returned output. Otherwise run the scripts with bash from the repository root.

Instructions

  1. Run scripts/snapshot.sh for the all-in-one JSON view (preferred default).

  2. For a quick human-readable memory line, run scripts/mem_summary.sh.

  3. To explain a single tegrastats line the user has pasted, see references/tegrastats-fields.md.

  4. To explain the NvMap clients output, see references/nvmap-clients.md.

Reporting guidance

Run the matching helper script before summarizing device state, and report only fields returned by that script. If direct execution is blocked by the runtime, run it with bash {baseDir}/scripts/<script-name> rather than trying to chmod files.

  • For "what is this Jetson" questions, quote product_model or sku, variant, l4t_version, and mem_total_gb.

  • For "slow and hot" questions, run snapshot.sh and summarize both sides of the symptom: thermal_c for heat, plus top_processes, gpu_processes, nvmap.top_clients, or gpu_source for load. End with a concrete handoff such as jetson-memory-audit, jetson-headless-mode, or jetson-inference-mem-tune.

  • For "which process is using memory" questions, run snapshot.sh and name the leading process as pid <number>, cmd, and its pss_kb / MiB value. If NvMap GPU memory is the relevant signal, also quote gpu_source and the top nvmap.top_clients or gpu_processes entry.

If your agent runtime does not automatically execute helper scripts relative to this skill directory, resolve script paths with the AgentSkills {baseDir} placeholder:

{baseDir}/scripts/snapshot.sh
{baseDir}/scripts/mem_summary.sh

Do not call jetson-diagnostic as a tool name unless the runtime explicitly registers skills as callable tools; Agent Skills are normally instructions plus files, not direct tool functions.

All scripts source the canonical platform detector at skills/jetson-diagnostic/scripts/detect_jetson.sh (exports JETSON_SKU, JETSON_GENERATION, JETSON_PRODUCT_LINE, JETSON_VARIANT, JETSON_MEM_GB, JETSON_L4T_VERSION, JETSON_PRODUCT_MODEL). Other skills may source this detector rather than duplicating Jetson identification logic. Exits 2 with a remediation message off-platform.

Limitations

  • Seeing this skill file does not guarantee access to Jetson host hardware. If /proc/device-tree/model, /etc/nv_tegra_release, tegrastats, nvpmodel, nvidia-smi, or /sys/kernel/debug/nvmap are missing inside a NemoClaw/OpenClaw sandbox, say the sandbox lacks Jetson host visibility and ask the user to run on the Jetson host or relaunch with a host-visible sandbox profile.

  • NvMap debugfs often requires root, so unprivileged runs may report gpu_source: "none" or incomplete nvmap fields.

  • This skill reports observed state only. Do not fabricate memory, GPU, thermal, service, or reclamation data when a tool is missing or inaccessible.

Error handling

  • If a helper exits off-platform, report that the current environment is not a Jetson host or lacks host visibility; do not substitute generic Linux values.

  • If tegrastats, nvpmodel, nvidia-smi, or NvMap debugfs are unavailable, preserve the corresponding null, false, or empty fields from the JSON and explain which signal is limited.

  • If snapshot.sh emits malformed JSON, report the raw failure and rerun after fixing the helper output; do not hand-edit a synthetic device snapshot.

Output contract for snapshot.sh

{
  "sku": "orin-nano",
  "generation": "orin",
  "product_line": "orin-nano",
  "variant": "orin-nano-8gb",
  "mem_total_gb": 8,
  "l4t_version": "36.4.0",
  "product_model": "nvidia jetson orin nano developer kit",
  "memory_kb": { "total": 8123456, "available": 4123456, "swap_total": 0, "swap_free": 0, "cached": 1234567 },
  "tegrastats_sample": "RAM 4011/8138MB (lfb 8x4MB) ...",
  "thermal_c": { "CPU": 52.3, "GPU": 49.0, "AO": 47.0 },
  "power": { "nvpmodel_id": 0, "nvpmodel_name": "MAXN" },
  "disk": [ { "mount": "/", "used_pct": 41 } ],
  "gpu_source": "nvmap:iovmm-clients",
  "gpu_devices": [],
  "gpu_processes": [],
  "nvmap": {
    "readable": true,
    "total_kb": 654321,
    "stats_total_bytes": 669985280,
    "top_clients": [ { "pid": 1234, "cmd": "vlm-server", "kb": 524288 } ]
  },
  "top_processes": [ { "pid": 4321, "cmd": "vllm", "pss_kb": 4000000 } ],
  "candidate_services": { "gdm3": { "active": "inactive", "enabled": "disabled" } }
}

gpu_source names the specific datum the skill used to attribute per-process GPU memory, so the caller can tell exactly what the numbers represent:

  • "nvidia-smi:compute-apps" — per-process used_memory values from nvidia-smi --query-compute-apps. Used on the unified nvidia.ko stack (Thor family today). Note: on this stack nvidia-smi's device-level memory.used query returns [N/A] on some BSPs, which is why the skill sums the per-process list rather than reading a top-level total. The summed total appears in gpu_processes[*].used_mib.

  • "nvmap:iovmm-clients" — per-process sizes from /sys/kernel/debug/nvmap/iovmm/clients. Used on the nvgpu stack (Orin family today), where nvidia-smi is a stub that returns [N/A] for every compute/memory query. Per-process entries appear in nvmap.top_clients; the kernel-side total is in nvmap.total_kb and (when readable) nvmap.stats_total_bytes.

  • "none" — no authoritative source reachable. Typical when running unprivileged on an nvgpu-stack Jetson (debugfs under /sys/kernel/debug/nvmap needs sudo); rerun with sudo to populate the nvmap fields.

The agent should present the salient parts back to the user (SKU, available memory, top GPU consumer per gpu_source, hottest zone, power mode) and offer to drill into specifics (top_processes, gpu_processes / nvmap, services).

Safety

This skill is read-only. It does not change nvpmodel, does not run jetson_clocks, does not modify services. To act on findings, hand off to:

  • jetson-memory-audit — focused memory snapshot + drop_caches verify loop

  • jetson-headless-mode — disable GUI + auxiliary daemons (safe, reversible)

  • jetson-inference-mem-tune — pick runtime + memory flags (vLLM / SGLang / llama.cpp / TensorRT Edge-LLM)

  • jetson-llm-serve — vLLM and related GHCR images with Jetson defaults

  • jetson-llm-benchmark — reproducible latency / throughput benchmarks

  • jetson-package — GHCR + Jetson AI Lab PyPI indexes vs generic ARM wheels

Cross-platform behavior

FamilyVariants the skill recognisestegrastatsnvidia-sminvpmodelNvMap debugfs
Jetson Thorthor-t5000, thor-t4000yesyes (full)yesyes (root)
Jetson AGX Orinorin-agx-64gb, orin-agx-32gb, orin-agx-industrialyesyes (stub, nvgpu)*yesyes (root)
Jetson Orin NXorin-nx-16gb, orin-nx-8gbyesyes (stub, nvgpu)*yesyes (root)
Jetson Orin Nanoorin-nano-8gb, orin-nano-4gbyesyes (stub, nvgpu)*yesyes (root)

* On Jetsons whose GPU is driven by the in-tree nvgpu kernel driver, the nvidia-smi binary is present but most fields (Memory-Usage, power, utilisation, compute-process table) report Not Supported / N/A. To decide which source to trust at runtime, the script does a capability probe — nvidia-smi --query-gpu=memory.used --format=csv,noheader,nounits — and only uses nvidia-smi for per-process GPU memory when that query returns a real integer. When it doesn't, the script falls back to /sys/kernel/debug/nvmap/iovmm/clients, which on nvgpu-stack Jetsons is the authoritative per-process GPU-memory source.

The script handles each tool's presence gracefully and reports null / false for tools it cannot reach (typical when the agent isn't running with the privilege needed for /sys/kernel/debug). Variant detection uses the /proc/device-tree/model string first (recognising names like T5000 / T4000) and falls back to memory-size heuristics when the model string is generic.

Files

scripts/snapshot.sh

#!/usr/bin/env bash
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# snapshot.sh
#
# All-in-one Jetson health snapshot. Read-only with respect to system state;
# uses `sudo -n` only to query privileged status when available.
#
# Usage:
#   snapshot.sh [--human] [--tegra-secs N] [--top-procs N]
#
# Outputs: JSON to stdout (default) or pretty-printed JSON with --human.
#
# Exit codes:
#   0  ok
#   2  not on a Jetson

set -uo pipefail

HUMAN=0
TEGRA_SECS=3
TOP_PROCS=10

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
# shellcheck source=common.sh
# shellcheck disable=SC1091
. "$SCRIPT_DIR/common.sh"

while [ $# -gt 0 ]; do
    case "$1" in
        --human)         HUMAN=1; shift ;;
        --tegra-secs)    need_value "$@"; TEGRA_SECS="$2"; shift 2 ;;
        --top-procs)     need_value "$@"; TOP_PROCS="$2"; shift 2 ;;
        -h|--help)
            grep '^#' "$0" | sed 's/^# \{0,1\}//'
            exit 0 ;;
        *) echo "unknown flag: $1" >&2; exit 64 ;;
    esac
done

# shellcheck source=detect_jetson.sh
# shellcheck disable=SC1091
. "$SCRIPT_DIR/detect_jetson.sh" || exit $?
# shellcheck source=runtime_probes.sh
# shellcheck disable=SC1091
. "$SCRIPT_DIR/runtime_probes.sh"

: "${JETSON_SKU:=unknown}"
: "${JETSON_GENERATION:=unknown}"
: "${JETSON_PRODUCT_LINE:=unknown}"
: "${JETSON_VARIANT:=unknown}"
: "${JETSON_MEM_GB:=null}"
: "${JETSON_L4T_VERSION:=unknown}"
: "${JETSON_PRODUCT_MODEL:=unknown}"

json_escape() {
    python3 -c 'import json,sys; print(json.dumps(sys.stdin.read().rstrip("\n")))' 2>/dev/null \
        || sed -e 's/\\/\\\\/g' -e 's/"/\\"/g' -e ':a;N;$!ba;s/\n/\\n/g'
}

# --- Memory ---
MEM_TOTAL_KB=$(read_meminfo_kb MemTotal)
MEM_AVAIL_KB=$(read_meminfo_kb MemAvailable)
MEM_FREE_KB=$(read_meminfo_kb MemFree)
SWAP_TOTAL_KB=$(read_meminfo_kb SwapTotal)
SWAP_FREE_KB=$(read_meminfo_kb SwapFree)
CACHED_KB=$(read_meminfo_kb Cached)

# --- Tegrastats sample ---
TEGRA_LINE=""
if command -v tegrastats >/dev/null 2>&1; then
    TEGRA_LINE=$(timeout "$TEGRA_SECS" tegrastats --interval 1000 2>/dev/null | tail -n 1)
fi
TEGRA_LINE_JSON=$(printf '%s' "$TEGRA_LINE" | json_escape)

# --- Thermal zones ---
THERMAL_PARTS=()
if [ -d /sys/class/thermal ]; then
    for zone in /sys/class/thermal/thermal_zone*; do
        [ -r "$zone/type" ] || continue
        [ -r "$zone/temp" ] || continue
        type=$(cat "$zone/type" 2>/dev/null)
        temp=$(cat "$zone/temp" 2>/dev/null)
        # Most Jetson zones report milli-degrees.
        c=$(awk -v t="$temp" 'BEGIN { printf "%.1f", t/1000.0 }')
        # Skip negative or absurd readings.
        case "$c" in -*|0.0) continue ;; esac
        THERMAL_PARTS+=("\"$type\":$c")
    done
fi
THERMAL_JSON="{$(IFS=,; echo "${THERMAL_PARTS[*]:-}")}"

# --- Power model ---
NVPMODEL_ID="null"
NVPMODEL_NAME="\"unknown\""
if command -v nvpmodel >/dev/null 2>&1; then
    out=$(sudo -n nvpmodel -q 2>/dev/null || nvpmodel -q 2>/dev/null || true)
    if [ -n "$out" ]; then
        id=$(printf '%s\n' "$out" | awk '/NV Power Mode/ {getline; print $1; exit}')
        name=$(printf '%s\n' "$out" | awk -F': ' '/NV Power Mode/ {print $2; exit}')
        [ -n "$id" ] && NVPMODEL_ID="$id"
        [ -n "$name" ] && NVPMODEL_NAME="\"$name\""
    fi
fi

# --- Disk ---
# df -P (POSIX): Filesystem 1024-blocks Used Available Capacity% Mounted-on
# $1=source $2=blocks $3=used $4=avail $5=capacity% $6=mount
DISK_PARTS=()
while IFS= read -r line; do
    mount=$(echo "$line" | awk '{print $6}')
    used=$(echo "$line" | awk '{print $5}' | tr -d '%')
    case "$mount" in
        ""|/dev|/dev/*|/proc|/sys|/run|/run/*|/snap/*|/sys/*|/boot/efi) continue ;;
        /var/lib/docker/*) continue ;;
    esac
    case "$used" in ''|*[!0-9]*) continue ;; esac
    DISK_PARTS+=("{\"mount\":\"$mount\",\"used_pct\":$used}")
done < <(df -P 2>/dev/null | tail -n +2)
DISK_JSON="[$(IFS=,; echo "${DISK_PARTS[*]:-}")]"

# --- GPU memory source selection ---
# On Jetsons using the `nvgpu` in-tree driver for CUDA (Orin family today), the
# `nvidia-smi` binary is present but most fields (Memory-Usage, memory.used,
# compute-apps) return [N/A] / "Not Supported" because the SMI CLI is bound to
# the display-only nvidia.ko module and cannot query the compute path.
# On Jetsons using the unified `nvidia.ko` compute driver (Thor family today),
# `nvidia-smi --query-compute-apps=used_memory` enumerates CUDA processes with
# real per-process sizes (though `--query-gpu=memory.used` still reports
# "Not Supported" on some Thor BSPs).
# `nvsmi_useful` lives in jetson-diagnostic/scripts/runtime_probes.sh: it returns true
# only when nvidia-smi reports a non-stub GPU name (no "(nvgpu)" suffix), the
# single capability probe that distinguishes the unified nvidia.ko stack from
# the legacy nvgpu stub-over-nvgpu stack across every L4T we have data for.

# Coerce a possibly non-numeric nvidia-smi field to either an integer or "null".
_nvsmi_int_or_null() {
    local v; v=$(printf '%s' "$1" | tr -d ' ')
    case "$v" in
        ''|*[!0-9]*) printf 'null' ;;
        *)           printf '%s' "$v" ;;
    esac
}

GPU_PROCESSES_JSON="[]"
GPU_DEVICES_JSON="[]"
# `gpu_source` is the *specific* datum used to attribute GPU memory:
#   "nvidia-smi:compute-apps"   per-process `used_memory` from
#                               `nvidia-smi --query-compute-apps` (unified
#                               nvidia.ko stack, e.g. Thor).
#   "nvmap:iovmm-clients"       per-process sizes from
#                               /sys/kernel/debug/nvmap/iovmm/clients
#                               (nvgpu stack, e.g. Orin).
#   "none"                      no source reachable (typically unprivileged
#                               invocation without sudo, and no nvidia-smi).
GPU_SOURCE="none"
if nvsmi_useful; then
    GPU_SOURCE="nvidia-smi:compute-apps"
    if NVIDIA_SMI_JSON=$(nvidia-smi \
        --query-compute-apps=pid,process_name,used_memory \
        --format=csv,noheader,nounits 2>/dev/null \
        | awk -F', *' '
            BEGIN { printf "["; first=1 }
            NF>=3 {
                pid=$1; mib=$NF;
                # Re-join any middle commas that the process name might contain.
                cmd="";
                for (i=2; i<NF; i++) cmd=(cmd=="" ? $i : cmd ", " $i);
                n=split(cmd, a, "/"); if (n>1) cmd=a[n];
                gsub(/"/, "\\\"", cmd);
                if (!first) printf ",";
                printf "{\"pid\":%d,\"cmd\":\"%s\",\"used_mib\":%d}", pid, cmd, mib;
                first=0;
            }
            END { printf "]" }
        ' 2>/dev/null); then
        GPU_PROCESSES_JSON="${NVIDIA_SMI_JSON:-[]}"
    else
        GPU_PROCESSES_JSON="[]"
    fi
    # memory.{total,used,free} report "Not Supported" / [N/A] on some BSPs even
    # when compute-apps work; emit null for any non-numeric field rather than
    # forcing a fake zero.
    if NVIDIA_SMI_JSON=$(nvidia-smi \
        --query-gpu=index,name,memory.total,memory.used,memory.free \
        --format=csv,noheader,nounits 2>/dev/null \
        | awk -F', *' '
            function intornull(s,   v) {
                v=s; gsub(/ /, "", v);
                if (v ~ /^[0-9]+$/) return v;
                return "null";
            }
            BEGIN { printf "["; first=1 }
            NF>=5 {
                gsub(/"/, "\\\"", $2);
                if (!first) printf ",";
                printf "%s%s%s%s%s%s%s%s%s%s%s",
                    "{\"index\":", $1, ",\"name\":\"", $2,
                    "\",\"memory_total_mib\":", intornull($3),
                    ",\"memory_used_mib\":", intornull($4),
                    ",\"memory_free_mib\":", intornull($5), "}";
                first=0;
            }
            END { printf "]" }
        ' 2>/dev/null); then
        GPU_DEVICES_JSON="${NVIDIA_SMI_JSON:-[]}"
    else
        GPU_DEVICES_JSON="[]"
    fi
fi

# --- NvMap (authoritative per-process GPU-memory source on nvgpu-stack Jetsons;
# harmless but near-empty on unified-driver Jetsons where nvidia-smi is used) ---
NVMAP_TOTAL_KB=0
NVMAP_TOP_JSON="[]"
NVMAP_STATS_TOTAL_BYTES="null"
NVMAP_PATH=/sys/kernel/debug/nvmap/iovmm/clients
NVMAP_STATS_PATH=/sys/kernel/debug/nvmap/stats/total_memory
NVMAP_READABLE=false
if [ -r "$NVMAP_PATH" ]; then
    NVMAP_READABLE=true
    # Only promote nvmap to the primary gpu_source if nvidia-smi didn't claim it.
    [ "$GPU_SOURCE" = "none" ] && GPU_SOURCE="nvmap:iovmm-clients"
    # iovmm/clients layout (all Jetsons today):
    #   CLIENT                        PROCESS      PID        SIZE
    #   user                  VLLM::EngineCor  1792595   14617792K
    #   ...
    #   total                                              15319584K
    # The SIZE column is documented as bytes with a K/M (and historically G)
    # suffix. Today's kernels emit K, but we normalize to KB defensively so
    # an M/G value cannot cause a 1024x under-count.
    NVMAP_TOTAL_KB=$(awk '$1=="total" {
            s=$NF;
            if      (s ~ /G$/) { gsub("G","",s); s=s*1024*1024 }
            else if (s ~ /M$/) { gsub("M","",s); s=s*1024 }
            else               { gsub("K","",s) }
            print s; exit
        }' "$NVMAP_PATH" 2>/dev/null)
    case "$NVMAP_TOTAL_KB" in ''|*[!0-9]*) NVMAP_TOTAL_KB=0 ;; esac
    NVMAP_TOP_JSON=$(awk '
            NR==1 { next }
            $1=="total" { exit }
            NF>=4 && $3 ~ /^[0-9]+$/ {
                pid=$3; cmd=$2; size=$NF;
                if      (size ~ /G$/) { gsub("G","",size); size=size*1024*1024 }
                else if (size ~ /M$/) { gsub("M","",size); size=size*1024 }
                else                  { gsub("K","",size) }
                if (size+0 == 0) next;
                printf "%d\t%s\t%s\n", size, pid, cmd;
            }
        ' "$NVMAP_PATH" 2>/dev/null \
        | sort -k1 -nr \
        | head -n "$TOP_PROCS" \
        | awk '
            BEGIN { printf "["; first=1 }
            {
                gsub(/"/, "\\\"", $3);
                if (!first) printf ",";
                printf "{\"pid\":%s,\"cmd\":\"%s\",\"kb\":%s}", $2, $3, $1;
                first=0;
            }
            END { printf "]" }
        ' 2>/dev/null || echo "[]")
    [ -z "$NVMAP_TOP_JSON" ] && NVMAP_TOP_JSON="[]"
fi
if [ -r "$NVMAP_STATS_PATH" ]; then
    v=$(cat "$NVMAP_STATS_PATH" 2>/dev/null)
    case "$v" in [0-9]*) NVMAP_STATS_TOTAL_BYTES="$v" ;; esac
fi

# --- Top PSS processes ---
TOP_PROC_JSON="[]"
if command -v procrank >/dev/null 2>&1; then
    # Capture sudo output separately so pipefail doesn't confuse sudo's exit
    # code (1 = password required) with awk's, which would cause "|| echo"
    # to fire and produce the invalid "[][]" double-array.
    _procrank_out=$(sudo -n procrank 2>/dev/null) || true
    TOP_PROC_JSON=$(printf '%s\n' "$_procrank_out" | awk -v n="$TOP_PROCS" '
        BEGIN {first=1; c=0; printf "["}
        NR>1 && NF>=6 && c<n {
            pid=$1; pss=$4; cmd=$NF;
            gsub("K","",pss);
            if (!first) printf ",";
            printf "{\"pid\":%s,\"pss_kb\":%s,\"cmd\":\"%s\"}", pid, pss, cmd;
            first=0; c++;
        }
        END {printf "]"}
    ')
else
    # Fallback: smaps_rollup walk. Slower but no extra binary needed.
    TOP_PROC_JSON=$(python3 - "$TOP_PROCS" <<'PY' 2>/dev/null || echo "[]"
import json, os, sys
n = int(sys.argv[1])
rows = []
for pid in os.listdir("/proc"):
    if not pid.isdigit():
        continue
    try:
        with open(f"/proc/{pid}/smaps_rollup") as f:
            pss_kb = 0
            for line in f:
                if line.startswith("Pss:"):
                    pss_kb = int(line.split()[1])
                    break
        with open(f"/proc/{pid}/comm") as f:
            comm = f.read().strip()
        rows.append({"pid": int(pid), "pss_kb": pss_kb, "cmd": comm})
    except (OSError, ValueError):
        continue
rows.sort(key=lambda r: r["pss_kb"], reverse=True)
print(json.dumps(rows[:n]))
PY
)
fi

# --- Service inventory (subset commonly relevant) ---
# Note: `systemctl is-active`/`is-enabled` for unit aliases or not-found units can
# emit two tokens ("not-found\ndisabled"); keep only the first line so the JSON
# values stay single-token.
SERVICES=(gdm3 gdm lightdm sddm display-manager pulseaudio bluetooth ModemManager \
          cups cups-browsed snapd whoopsie kerneloops avahi-daemon \
          unattended-upgrades packagekit nvargus-daemon containerd docker)
SVC_PARTS=()
for svc in "${SERVICES[@]}"; do
    state=$(systemctl is-active "$svc" 2>/dev/null) || true; state="${state%%$'\n'*}"
    enabled=$(systemctl is-enabled "$svc" 2>/dev/null) || true; enabled="${enabled%%$'\n'*}"
    : "${state:=inactive}"; : "${enabled:=disabled}"
    SVC_PARTS+=("\"$svc\":{\"active\":\"$state\",\"enabled\":\"$enabled\"}")
done
SVC_JSON="{$(IFS=,; echo "${SVC_PARTS[*]}")}"

DEFAULT_TARGET=$(systemctl get-default 2>/dev/null || echo unknown)

# --- Emit JSON ---
PRODUCT_MODEL_JSON=$(printf '%s' "$JETSON_PRODUCT_MODEL" | sed -e 's/\\/\\\\/g' -e 's/"/\\"/g')

PAYLOAD=$(cat <<EOF
{
  "sku": "${JETSON_SKU}",
  "generation": "${JETSON_GENERATION}",
  "product_line": "${JETSON_PRODUCT_LINE}",
  "variant": "${JETSON_VARIANT}",
  "mem_total_gb": ${JETSON_MEM_GB},
  "l4t_version": "${JETSON_L4T_VERSION}",
  "product_model": "${PRODUCT_MODEL_JSON}",
  "default_systemd_target": "${DEFAULT_TARGET}",
  "memory_kb": {
    "total": ${MEM_TOTAL_KB:-0},
    "available": ${MEM_AVAIL_KB:-0},
    "free": ${MEM_FREE_KB:-0},
    "cached": ${CACHED_KB:-0},
    "swap_total": ${SWAP_TOTAL_KB:-0},
    "swap_free": ${SWAP_FREE_KB:-0}
  },
  "tegrastats_sample": ${TEGRA_LINE_JSON:-"\"\""},
  "thermal_c": ${THERMAL_JSON},
  "power": { "nvpmodel_id": ${NVPMODEL_ID}, "nvpmodel_name": ${NVPMODEL_NAME} },
  "disk": ${DISK_JSON},
  "gpu_source": "${GPU_SOURCE}",
  "gpu_devices": ${GPU_DEVICES_JSON},
  "gpu_processes": ${GPU_PROCESSES_JSON},
  "nvmap": {
    "readable": ${NVMAP_READABLE},
    "total_kb": ${NVMAP_TOTAL_KB:-0},
    "stats_total_bytes": ${NVMAP_STATS_TOTAL_BYTES},
    "top_clients": ${NVMAP_TOP_JSON}
  },
  "top_processes": ${TOP_PROC_JSON},
  "candidate_services": ${SVC_JSON}
}
EOF
)

if [ "$HUMAN" = "1" ] && command -v python3 >/dev/null 2>&1; then
    printf '%s' "$PAYLOAD" | python3 -m json.tool
else
    printf '%s\n' "$PAYLOAD"
fi

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Security

What our scanner noticed· Medium risk

  • Large base64-encoded blobs
  • Scripts that your agent may run

This skill includes scripts that your agent may run. Read them before you use it.

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Created by
Jetson Team
Publisher
NVIDIA (Official)
Distributed by
NVIDIA
Source
NVIDIA/skills · skills/jetson-diagnostic · commit fce3cca
Licence
Apache License 2.0 (redistribution allowed)
  • LICENSE (added by AutoGineer from LICENSE-APACHE)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by Jetson Team or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download jetson-diagnostic.zip