All skills
nvidia avatar

/jetson-customize-camera

@87053a0
by NVIDIA Corporationnvidia/skills3.5k stars
427

Enable MIPI/GMSL camera sensors on a Jetson Thor or Orin custom carrier by rendering a kernel-DT overlay from the in-tree sensor DTSI. Do NOT use for UPHY lane allocation or ODMDATA edits.

Use this Skill: https://skilld.dev/gh/nvidia/skills/jetson-customize-camera

This session only. Nothing lands on disk.

referencescamera-overlay-templatesREADME.md

≈661 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Camera overlay templates

These are agent-consumable starting templates for the kernel-DT overlays /jetson-modify-camera emits. The skill is agentic — there is no Python renderer; the agent picks the closest template by camera type, decompiles a matching in-tree reference overlay for the sensor-specific binding details, and patches placeholders.

How to pick a template

Camera class Template When to use In-tree exemplar (decompile for sensor specifics)
DPHY-direct (single sensor on a CSI port, no SerDes) dphy-direct.dts.tmpl imx274, imx185, imx390, e3331, e3333, eCAM130A, ar0234 (direct), and any future sensor whose datasheet says "MIPI CSI-2 D-PHY" with no serializer tegra264-p3971-camera-<sensor>-overlay.dtbo
GMSL serdes chain (sensor → MAX9295 serializer → coax → MAX96712 deserializer → CSI) gmsl-serdes.dts.tmpl Hawk (AR0234 over GMSL), Owl (OX03A10 over GMSL), Leopard p3762 daughtercards, anything with MAX96712 / MAX9296 in the chain tegra264-p3971-camera-p3762-a00-<config>-overlay.dtbo

Other GMSL deserializers (MAX9296 standalone, etc.) follow the same GMSL template — the deserializer compatible string and channel layout change but the structural pattern (vi → nvcsi → tca9546 mux → deserializer → serializer → sensor) is identical.

Placeholder grammar

Every placeholder is <<UPPER_SNAKE_CASE>> so a simple sed -e "s|<<KEY>>|value|g" (or the agent's Edit tool) substitutes it. The skill's session.camera block names every placeholder it sets — see SKILL.md §"Output contract".

How the agent uses these templates

Per /jetson-modify-camera SKILL.md step 6:

  1. Read the template DTS that matches the chosen camera class.
  2. Also decompile the closest in-tree tegra<soc>-p3971-camera-*-overlay.dtbo for the user's exact sensor model — that's the canonical source for sensor mode-table entries (mode_type, pixel_phase, sensor-mode reg-list rows). Splice those into the sensor@<addr> subnode of the template.
  3. Substitute the placeholders from session.camera.
  4. Write $L4T_DIR/kernel/dtb/tegra<soc>-overlay-<carrier_short>-camera.dts and compile to .dtbo.
  5. Append to OVERLAY_DTB_FILE in the carrier .conf after the platform *-dynamic.dtbo.

The templates intentionally cover the structure (VI/NVCSI port plumbing, mux subnode shape, sensor-driver compatible slot). Sensor-specific register tables come from the in-tree exemplar — they encode silicon constraints that hand-coding would silently miss.

Source: SKILL.md on GitHub

1 warning3mo3 checks · Risk SAFE
  • Gen Agent Trust Hub3mo

    The jetson-customize-camera skill is a specialized engineering utility from NVIDIA for configuring MIPI/GMSL camera sensors on Jetson platforms. It automates the generation of Device Tree Overlays by using standard Linux development tools like the Device Tree Compiler (dtc), C Preprocessor (cpp), and Git. The analysis confirms that the skill performs legitimate technical tasks within a local development environment and shows no evidence of malicious patterns, data exfiltration, or unauthorized privilege escalation.

  • Socket3mo

    No alerts

  • Snyk3mo

    Risk: MEDIUM · 1 issue

Signed by skilld at 87053a0. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 1 hour ago.

Activeupdated 3 months ago
version
0.0.1
Other metadata
metadata
{
  "data-classification": "public",
  "author": "Jetson Team",
  "tags": [
    "bsp",
    "phase-2",
    "io",
    "camera",
    "csi"
  ],
  "domain": "meta"
}

README badge

README badge for nvidia/skills/jetson-customize-camera