skilld
All skills
nvidia avatar

/jetson-video-setup

@9bb5a39
by NVIDIA Corporationnvidia/skills3.5k stars
424

Use when installing, repairing, reusing, inspecting, or verifying readiness of the native NVIDIA Video Codec SDK or PyNvVideoCodec on Jetson, including the one-frame encode/decode smoke test with official samples, and when interpreting what those readiness results, including CPU-buffer and device-memory sample modes, do and do not establish.

Ask your Agent

Use this Skill: https://skilld.dev/gh/nvidia/skills/jetson-video-setup

This session only. Nothing lands on disk.

referencesvideo-content.md

≈1.3k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Video content policy

This setup-owned reference defines the shared media-selection and evidence policy for every Jetson video skill. Setup owns the policy and deterministic smoke fixture; benchmark and pipeline workflows own normal media execution.

Route matrix

Workflow Content rule
Setup/readiness verification A deterministic synthetic raw fixture is allowed; user media is not required.
Capability API query Media-free. A minimal capability-operation smoke test may use setup's deterministic fixture.
Recipe planning, validation, or live compatibility check Media-free. Stop before media work.
Documentation-only benchmark estimate Media-free. Do not execute a benchmark operation or claim a measurement.
Live benchmark encode, decode, compare, or camera_capacity Require exact user-selected media under the input gate below.
Pipeline encode_decode, native_transcode, pynvc_segments, container_triage, av1_verify, or acceptance Require exact user-selected media under the input gate below.
Direct recipe-bound encode/decode execution Require exact user-selected media, except for the capability-smoke fixture above.
Pipeline content-summary validation Consume the already bound external media and evidence; never select or acquire replacement media.

Never use synthetic, uniform, black, or generated content for recipe execution, performance/benchmarking, a content-sensitive comparison, or a pipeline.

Input gate

Recipe planning and validation, capability queries, and documentation-only benchmark estimates are media-free. Before any recipe execution, live benchmark dry run or execution, content-sensitive comparison, or pipeline dry run or execution, require the user to supply exactly one of:

  • an absolute target-local media path; or
  • one exact HTTP(S) media URL.

If neither is present, return input_required with next_action: provide_media_path_or_url, ask the user for one, and pause before probing, retrieving, converting, preparing an operation invocation, or launching an operation. Never browse for, select, recommend, or retrieve media from a bundled catalog or an alternate source. The synthetic-content prohibition is an agent-routing rule, not a byte-content classifier: an artifact identity alone does not prove who selected the file or how its frames were created. Bind the operation to the user's supplied path or URL and never turn the setup fixture into an execution input.

Treat user selection as operation input, not as proof of copyright or license rights. Preserve license and attribution exactly when evidenced or supplied. Use the literal string unknown when either is unknown; never infer or invent it.

Resolve the selected source

  • For a local path, require a canonical absolute regular-file path on the target, keep the source read-only, and record its current size and SHA-256. Set source_url to JSON null.
  • For a URL, first prove validated target eligibility and an authenticated released route for the intended operation. Retrieve only the exact user-supplied HTTP(S) URL into the fresh workspace. Record that exact requested URL as source_url and preserve the retrieval command evidence. Never substitute a mirror, nearby catalog entry, or different agent-selected URL without asking the user.
  • If the source is missing, unreadable, empty, changes identity, or cannot be retrieved, report the evidenced input failure and ask for a replacement. Do not relabel it unsupported or operation_failed.

Use in workflows

  • Classify the selected media as raw, elementary-stream, or container input before choosing a released sample. Invoke only the authenticated, release-owned NVIDIA sample selected by the workflow for demux, decode, or raw-frame production.
  • For an encoder run, create raw frames only through an approved NVIDIA sample that can write the required format. Record the source and derived SHA-256, geometry, format, frame rate, frame count, and decode evidence. Do not pass a container file to a raw-input runner.
  • If the selected NVIDIA sample does not accept the user-selected source, ask for another local path or URL, or report the route blocked. Do not choose fallback media or use FFmpeg to replace codec work.
  • Use the same user-selected content and frame range across comparisons. In each owning workflow, the agent validates byte identity, frame alignment, metadata, and official-sample evidence; it does not infer scene complexity, representativeness, ownership, or nonuniformity. Never promote the setup fixture into a performance input. Report results only for the exact user-selected workload and do not generalize it into a complete workload suite.

Evidence

For every selected input, preserve source_url (the exact user URL, or JSON null for local media), honest license and attribution strings (the literal unknown is allowed), canonical path, byte size, observed SHA-256, codec/container, dimensions, frame rate, pixel format, bit depth, and frame count. Preserve a published source checksum only when it is actually available.

Source: SKILL.md on GitHub

Third-party checks

No alerts6d3 checks · Risk SAFE
  • Gen Agent Trust Hub6d

    A safe and highly controlled setup skill from NVIDIA for installing and verifying video SDKs on Jetson devices. It employs best practices such as cryptographic hash verification for all external downloads, strict validation of system configuration, and restricted use of privilege escalation.

  • Socket6d

    No alerts

  • Snyk6d

    Risk: LOW · No issues

Provenance

Signed by skilld at 9bb5a39. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last week

Capability

Other metadata
metadata
{
  "author": "Vinit Bansal <vinitkumarb@nvidia.com>",
  "tags": [
    "jetson",
    "video-codec-sdk",
    "pynvvideocodec",
    "setup",
    "nvenc",
    "nvdec"
  ],
  "languages": [
    "markdown"
  ],
  "data-classification": "public"
}

README badge

README badge for nvidia/skills/jetson-video-setup