cudaq-guide
CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.
jetson-download-bsp
Download NVIDIA Jetson Linux BSP artifacts (BSP tarball, sample rootfs, public_sources, x-tools, guides) for the active target. Use for Auto Setup; not for extraction or profile edits.
Full skill instructions
jetson-download-bsp owns network fetches for the Auto Setup path.
It does not extract archives, initialize source trees, register documents,
or edit the active target profile. Those remain owned by
jetson-init-image, jetson-init-source, and jetson-link-docs.
/jetson-quick-start selected Auto Setup.Resolve the active profile per
../../context/target-platform-contract.md.
Refuse if no active profile exists.
Follow the shared
quick_start_prefill contract.
When invoked by /jetson-quick-start, this skill consumes
quick_start_prefill.mode and quick_start_prefill.download.bsp_release
as the requested release when present. Missing or invalid prefilled values
are not fatal, but a concrete release is required before any download
starts.
Use the workspace root as the parent of target-platform/. Place files
where downstream setup skills already search:
| Artifact | Destination |
|---|---|
| Jetson Linux BSP tarball | <bsp_image.root_path or workspace/Image>/Jetson_Linux_R<ver>_aarch64.tbz2 |
| Sample rootfs tarball | <bsp_image.root_path or workspace/Image>/Tegra_Linux_Sample-Root-Filesystem_R<ver>_aarch64.tbz2 |
public_sources.tbz2 | <workspace>/Downloads/public_sources.tbz2 |
x-tools.tbz2 | <workspace>/Downloads/x-tools.tbz2 |
| Developer Guide | <documents.root_path or workspace/Documents>/ |
| User Guide | <documents.root_path or workspace/Documents>/ |
Do not write bsp_image:, source:, or documents: blocks. The next
setup skills will register derived metadata after extraction or user
confirmation.
Fetch the release list from: https://developer.nvidia.com/embedded/jetson-linux-archive
From the archive and linked release pages, list concrete releases that
explicitly support the active platform. Let the user choose a concrete
Jetson Linux / L4T release token (R36.4.4, 36.4.4, 38.2,
38.2.1, 38.4, etc.); do not proceed with a family placeholder such
as R38.x. A concrete release token is an optional leading R followed
by two or three numeric dotted components. Preserve the token exactly as
the official page presents it; do not add .0, remove a patch component,
or otherwise normalize it. Sort release choices by dotted numeric order
descending (newest first), comparing the components that are present; keep
source-provided recommendations in descriptions.
Resolve quick_start_prefill.download.bsp_release as follows:
skip, blank, or a family placeholder (R38.x, 38.x): show
supported concrete releases and ask the user to choose one.Never auto-substitute the requested release with a supported release, even when there is only one supported option.
For unsupported or ambiguous prefill, ask with the shared
User input prompt style,
to choose exactly one click-to-select action:
pick_supported_release — user selects one supported concrete release.confirm_override — user confirms the incompatible release and enters
an override reason.cancel — stop with no downloads.Do not continue to "Locate release artifacts" until the selected release is concrete and supported, or the user explicitly confirms override. Record any override reason in the final summary. If the user cancels, stop with no downloads. Do not claim support unless the archive or linked release page explicitly lists the active platform.
For every artifact WebFetch in "Resolve target release" and "Locate release artifacts" (BSP, rootfs,
public_sources.tbz2, x-tools.tbz2, Developer Guide, User Guide),
try 3 attempts before marking it missing:
.html indexes.https://docs.nvidia.com/jetson/archives/r<ver>/.Only after all 3 fail, route to user-paste / skip. Never fabricate URLs.
Follow links from the selected release page. Never construct NVIDIA download URLs by string concatenation.
Required artifacts:
public_sources.tbz2.x-tools.tbz2 / Crosstool-NG Toolchain gcc matching the selected
BSP release.If a guide is not available as a direct file link, accept the release documentation URL as the artifact and record it in the final summary. If the toolchain is listed on a JetPack downloads page instead of the release page, follow the official page link and match the selected BSP release; do not synthesize the URL from a release number. Do not proceed until each required artifact is downloaded, already present at its destination, represented by a URL, or explicitly skipped by the user.
Before changing the workspace, print a plan with:
reference_devkit.name, optional
custom_carrier.name, and flash_config);curl.Default to Download all. Do not ask an extra confirmation when every required artifact is downloadable, already present for reuse, or a documentation URL to record. Print the plan and continue to disk-space verification.
Still stop and ask before side effects in any of these cases:
Sum Content-Length across artifacts to download (curl -sI <url>);
refuse if df -B1 --output=avail <workspace> headroom < sum × 1.2.
Skip URL-only artifacts.
Create destination directories as needed. For direct file links, use a
temporary .part path and rename only after success:
mkdir -p "<destination-dir>"
curl -fL --retry 3 --retry-all-errors --retry-delay 5 -o "<destination>.part" "<artifact-url>"
mv "<destination>.part" "<destination>"
If a destination file already exists and is non-empty, reuse it by default after showing its path. Ask before overwriting. If an artifact download fails due to 404, 401/403, EULA/browser authentication, or a network error, stop and show the artifact URL plus expected destination path. Ask the user to download the file manually, skip that artifact, or cancel. Do not synthesize alternate NVIDIA URLs.
Before finishing, verify the BSP and rootfs tarball filenames contain
the exact R<ver> token from the official artifact URL or filename
discovered in Step 2. <ver> may have two or three numeric components
(R38.2, R38.2.1, R38.4, etc.). Do not infer R38.2.0 from
R38.2, and do not assume that a selected two-component release maps to
a three-component tarball name:
Jetson_Linux_R<ver>_aarch64.tbz2Tegra_Linux_Sample-Root-Filesystem_R<ver>_aarch64.tbz2Verify public_sources.tbz2 is a bzip2 tarball when it exists locally:
file -b "<workspace>/Downloads/public_sources.tbz2" | grep -q "bzip2 compressed"
Verify x-tools.tbz2 the same way when it exists locally:
file -b "<workspace>/Downloads/x-tools.tbz2" | grep -q "bzip2 compressed"
Warn, but do not fail, if document filenames differ from the release version; NVIDIA documentation naming is less consistent than tarballs.
Report the selected release, downloaded/reused/skipped artifacts, and exact local paths. Then hand back to the caller with:
/jetson-init-image/jetson-init-source/jetson-link-docsjetson-init-image and jetson-init-source.Downloads/, because
jetson-init-image searches the image root, workspace root, and
current directory.public_sources.tbz2 in Downloads/ because
jetson-init-source auto-discovers that path for Branch A.x-tools.tbz2 in Downloads/ because jetson-init-source
auto-extracts that path and writes the resulting source.toolchain
prefix.documents.bsp_user_guide in the profile
schema (alongside bsp_developer_guide). jetson-link-docs writes
it when the user confirms; this skill only stages the file or URL.Stage the four NVIDIA-shipped artifacts (BSP, sample rootfs,
public_sources, x-tools) plus the Developer / User Guides at the exact
paths the Setup skills look in, so /jetson-init-image,
/jetson-init-source, and /jetson-link-docs can run offline once
this skill is done.
../../context/target-platform-contract.md.developer.nvidia.com and the Jetson Linux archive
page.Image/, Source/,
Downloads/, and Documents/ roots (default workspace shape).apply_binaries.sh, mounts
sources, registers documents, or mutates the active profile./jetson-set-target.public_sources.tbz2 missing on archive — some releases ship it
as a separate page; follow the release-page link rather than
guessing the URL.CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.
Use when the user wants to create a dataset, generate synthetic data, or build a data generation pipeline.
NVIDIA DeepStream SDK development with Python pyservicemaker API. Use when building video analytics pipelines, GStreamer-based video processing, TensorRT inference integration, object detection/tracking, or Kafka/message broker integration.
Filesystem RAG benchmarks: corpus/, train.json, evaluate_rag.py (RAGAS quality). Not for prod monitoring, latency/throughput benchmarking (use rag-perf), or evals outside this repo layout.
Use when the user wants to search, query, extract, transcribe, describe, quote, filter, or aggregate across documents — PDFs, scanned forms / images (`.jpg` `.png` `.tiff`), Office (`.docx` `.pptx`), text (`.html` `.txt`), audio (`.mp3` `.wav` `.m4a`), or video (`.mp4` `.mov`). Prefer this over native Read / Grep for multi-file or non-PDF corpora. Not for: editing files, web browsing, single-file plain-text lookups, fine-tuning.
Coordinate the end-to-end CAD/source-asset to SimReady workflow. Use for broad requests such as CAD to SimReady, source asset to simulation-ready USD, or prop packaging that require conversion, material/physics assignment, SimReady conformance, validation, and optional package creation; deploy or verify Content Agents services first when property assignment is enabled; route single-stage work through nested references.
Use this skill to bring any vision model from HuggingFace or NVIDIA NGC into an NVIDIA DeepStream pipeline with end-to-end automation: ONNX download, SafeTensors export, TRT engine build, custom nvinfer bbox parser, multi-stream benchmark, and PDF report. Object detection models only.
Top-level workflow skill for USD performance diagnosis and optimization. Handles slow loading, high memory, low FPS, and broad scene-optimization requests; delegates auth/runtime setup to Phase 0 owners.
Vehicle routing (VRP, TSP, PDP) with cuOpt — Python API only. Use when the user is building or solving routing in Python.
Use only to generate or update a governance skill card for a specified existing agent skill directory. Do not use for explaining, listing, comparing, or discussing skill capabilities.
Use as the top-level router for Omniverse Realtime Viewer USD app requests and focused viewer reference documents.
Performance benchmarking for a deployed NVIDIA RAG Blueprint server: profiling pass + aiperf load test driven by a single YAML config. Not for accuracy / RAGAS scoring (use rag-eval) or for deploying / repairing services (use rag-blueprint).
Generate, edit, and beat-sync AI video with leading models in one workspace.
The world's fastest calendar for remote work
Transform Your Design with AI Designer by ImgCreator.ai
Revolutionizing Video Production with AI-Powered Creativity
Extend an image past the frame and let AI fill the new aspect ratio.
Discover your celebrity doppelgänger with StarByFace!
ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.
Opus.ai: Revolutionize Your Web Experience