Install the "jetson-quick-start" agent skill from https://github.com/NVIDIA/skills/tree/main/skills/jetson-quick-start into .claude/skills/jetson-quick-start/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "jetson-quick-start", then confirm the skill loads.
Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Type this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
skills CLI
$ npx skills add NVIDIA/skills --skill jetson-quick-start -a codex
Project install goes to .agents/skills/; add -g for ~/.codex/skills/.
Install the "jetson-quick-start" agent skill from https://github.com/NVIDIA/skills/tree/main/skills/jetson-quick-start into .agents/skills/jetson-quick-start/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "jetson-quick-start", then confirm the skill loads.
Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
skills CLI
$ npx skills add NVIDIA/skills --skill jetson-quick-start -a cursor
Project install goes to .agents/skills/; add -g for ~/.cursor/skills/.
Install the "jetson-quick-start" agent skill from https://github.com/NVIDIA/skills/tree/main/skills/jetson-quick-start into .cursor/skills/jetson-quick-start/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "jetson-quick-start", then confirm the skill loads.
Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
skills CLI
$ npx skills add NVIDIA/skills --skill jetson-quick-start -a gemini-cli
Project install goes to .agents/skills/; add -g for ~/.gemini/skills/.
Install the "jetson-quick-start" agent skill from https://github.com/NVIDIA/skills/tree/main/skills/jetson-quick-start into .gemini/skills/jetson-quick-start/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "jetson-quick-start", then confirm the skill loads.
Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Installs for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
skills CLI
$ npx skills add NVIDIA/skills --skill jetson-quick-start -a github-copilot
Project install goes to .agents/skills/; add -g for ~/.copilot/skills/.
Install the "jetson-quick-start" agent skill from https://github.com/NVIDIA/skills/tree/main/skills/jetson-quick-start into .github/skills/jetson-quick-start/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "jetson-quick-start", then confirm the skill loads.
GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
skills CLI
$ npx skills add NVIDIA/skills --skill jetson-quick-start -a opencode
OpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
Install the "jetson-quick-start" agent skill from https://github.com/NVIDIA/skills/tree/main/skills/jetson-quick-start into .opencode/skills/jetson-quick-start/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "jetson-quick-start", then confirm the skill loads.
OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
Skill name
jetson-quick-start
GitHub stars
3.5k
Token cost
~4.7k tokens
SKILL.md length
2,116 words
Files
5
Skills in repo
380
Repo updated
First seen
Licence
Apache-2.0
At a glance
Entry skill for Jetson / IGX BSP customization. An agent skill from NVIDIA/skills.
Works in 3 steps: Auto Setup → Guided Setup → Use Existing Workspace
Tasks that involve GPU and accelerator computing
SKILL.md covers Modes, Procedure, Purpose and Prerequisites, plus 4 more sections
Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md
What it does
Jetson Quick Start is an agent skill from NVIDIA/skills, published by the product's own GitHub organization. Entry skill for Jetson / IGX BSP customization. Asks one core click-to-select setup questionnaire and passes prefilled answers to downstream setup skills.
Its SKILL.md is about 4.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 5 other files (for example `BENCHMARK.md`, `evals/evals.json` and `skill-card.md`).
It sits in AI & LLM Engineering, covering GPU and accelerator computing. It works with NVIDIA AI Platform. The repository describes itself as: Agent Skills for NVIDIA products — install into Claude Code, Codex, and other coding agents to run Physical AI, robotics, simulation, CUDA, and RAG workflows end to end. The licence is Apache-2.0.
When your agent uses it
Tasks that involve GPU and accelerator computing
Example prompts
“/jetson-quick-start”
Workflow steps
3 steps, taken from the first numbered list in SKILL.md.
1Auto Setup
2Guided Setup
3Use Existing Workspace
What it can do on your machine
Read from SKILL.md and the folder at commit 0e0d506. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
No scripts in the folder and no shell commands in SKILL.md (its code samples are yaml).
From the folder's file list and the shell code blocks in SKILL.md.
Network
Links to these hosts (documentation or services it may open):
developer.nvidia.com
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Jetson Quick Start loads about 4.7k tokens when it runs. Until then it costs about 43 tokens; SKILL.md has 2,116 words of instructions outside code blocks.
Always· name and description, kept in context so the agent knows when to use it
~43
When it runs· the whole SKILL.md, loaded when a task matches
~4.7k
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check passed
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
Download SKILL.mdSave it as .claude/skills/jetson-quick-start/SKILL.md (or your agent's skills folder). This skill also uses 4 other files; get the full folder from GitHub.
name
jetson-quick-start
description
Entry skill for Jetson / IGX BSP customization. Asks one core click-to-select setup questionnaire and passes prefilled answers to downstream setup skills.
version
0.0.1
license
Apache-2.0
metadata.data-classification
public
metadata.author
Jetson Team
metadata.tags
setup, bootstrap, download
metadata.domain
setup
Quick Start — BSP customization entrypoint
jetson-quick-start is a dispatcher and intake form. It does not download,
extract, materialize sources, bind documents, or derive carrier files. It
chooses the setup path, gathers core answers, then hands those answers to
the skills that own the actual work.
Modes
Do not ask setup mode in a separate prompt. Include it as the first field
in the core quick-start questionnaire, with choices shown in this fixed
order:
Auto Setup
Guided Setup
Use Existing Workspace
Offer cancel separately as an exit action, not as a setup mode.
BSP image, source tree, and docs are already prepared
target setup, workspace verification, /jetson-generate-kb, optional /jetson-derive-carrier; route to init skills only for missing prerequisites
Target setup means keep the active profile, switch via /jetson-set-target,
or create one via /jetson-init-target.
Procedure
Print Disclaimer
Before opening the core questionnaire or inspecting setup inputs, print
this disclaimer block exactly once:
text
================================================================================
DISCLAIMER
These skills help automate Jetson BSP setup and customization, but they do not
replace NVIDIA official documentation or engineering review. Review generated
plans, commands, diffs, and commit messages before accepting them.
Flashing can erase device storage or leave a target temporarily unbootable.
Keep backups and verify the active target, BSP release, and hardware setup
before deploy steps.
================================================================================
Disclaimer Acceptance Gate
Immediately after printing the disclaimer, open a separate one-question
AskUserQuestions disclaimer form and wait for the user's submitted
answer before doing anything else. Do not collect disclaimer acceptance
with a plain-text chat reply, inferred default, prior run state, active
profile, cached answer, or any other non-UI signal.
The form must contain exactly one question:
accept_disclaimer — prompt: "Accept the disclaimer and continue
quick-start?" Choices: accept_disclaimer and cancel.
Treat only a submitted accept_disclaimer choice as acceptance. The
disclaimer form is separate from the core questionnaire and does not
replace any of the four routing-critical core questions.
If the form cannot be opened, is unavailable, is cancelled, returns no
submitted answer set, or returns any answer other than
accept_disclaimer, stop immediately and report that quick-start cannot
continue until the disclaimer is accepted through AskUserQuestions.
Mandatory Questionnaire Gate
After the disclaimer has been accepted, jetson-quick-start MUST open the
core AskUserQuestions questionnaire and wait for the user's submitted
answers before doing any setup dispatch. If the questionnaire cannot be
opened, is unavailable, is cancelled, or returns no submitted answer set,
stop immediately and report that quick-start cannot continue without the
questionnaire.
Do not continue by using existing active profiles, workspace state,
tarball names, cached answers, release compatibility, or "obvious"
defaults. Do not call downstream skills until this gate has a submitted
mode answer and normalized quick_start_prefill.
Survey State
Inspect target-platform/active_target.yml and existing
target-platform/*.yaml profiles. If an active profile exists, show
reference_devkit.name, optional custom_carrier.name, and active
flash_config. Do not infer platform identity from tarballs, repo names,
or document titles.
Read ../../references/bsp-platforms-catalogue.md
and parse every row in the product / chip / SKU / flash-config table.
This catalogue is the source for the full active_platform
questionnaire list. If the list is incomplete, update the catalogue;
do not duplicate or patch the platform list inside quick-start. Do not read
jetson-init-target/SKILL.md or run the init-target flow during
quick-start preprocessing; downstream target creation remains owned by
jetson-init-target.
Keep this survey lightweight so the core questionnaire appears quickly:
read target profile pointers/summaries, perform cheap path existence
checks when useful, and fetch only lightweight Jetson Linux release
metadata from the official archive needed to populate bsp_release
choices. Parse all current release sections, not only the section that
appears to match the selected platform. Do not scan large BSP / source /
document trees, inspect archives, generate KB content, or run setup tools
before the user submits the core questionnaire. Prepare the possible plans
for all three modes; choose the actual downstream plan only after the form
is submitted.
Print Platform Reference List
Before opening the AskUserQuestions UI, print the full platform list
parsed from bsp-platforms-catalogue.md with stable numeric indexes for
this run:
text
Available platforms:
T234 - Orin
1. Jetson AGX Orin 64GB | CVM P3701-0005 | CVB P3737-0000 | jetson-agx-orin-devkit.conf
...
T264 - Thor
N. Jetson AGX Thor T5000 | CVM P3834-0008 | CVB P4071-0000 | jetson-agx-thor-devkit.conf
Use these indexes only for the immediately following questionnaire.
Do not auto-select an index, even when there is an active profile or a
single obvious match.
Ask Core Questionnaire
Generate one click-to-select core questionnaire for the whole setup run
using the AskUserQuestions UI. Do not collect quick-start answers with
plain-text prompts, inline chat questions, inferred defaults, or manual
summary confirmation.
This is a hard gate: do not skip it, do not use a one-profile shortcut,
and do not auto-select mode, active_platform, bsp_release, or
custom_carrier from active_target.yml, filenames, tarballs, repo
names, release compatibility, workspace state, cached answers, or prior
runs. An existing active profile may be shown as an option, but it is not
selected until the user submits the AskUserQuestions form.
Because AskUserQuestions can ask at most four questions, the core form
must contain only routing-critical fields:
mode — Auto Setup, Guided Setup, Use Existing Workspace, or
cancel.
active_platform — use the printed full platform list; follow the
active-platform question rules below.
bsp_release — recent global concrete release candidate shortcuts
generated from the official Jetson Linux archive for this run, sorted
newest first by dotted numeric order, plus skip. Preserve each
official release token exactly as listed; concrete tokens may have two
or three numeric components, such as 38.2, 38.2.1, 38.4, or
R36.4.4. Do not normalize 38.2 to 38.2.0, and do not infer a
missing patch component. Do not hard-code or cache candidates. If
metadata fetch is slow or fails, still show skip and a typed concrete
release option. Do not platform-filter this list in quick-start; include
at least the newest concrete row from each current major release line
found in the archive. If parsing returns only one major release line,
treat the release list as incomplete and fall back to skip plus typed
input instead of showing a partial candidate list.
custom_carrier — if the selected/active profile already has a
custom carrier, offer keep_existing_custom_carrier,
no_custom_carrier, add_custom_carrier, and skip; otherwise offer
no_custom_carrier, add_custom_carrier, and skip.
The user submits this AskUserQuestions form once. Only after that
submission may mode, active platform, BSP release, and custom-carrier
intent decide which downstream skills run and which prefilled answers are
consumed or ignored. If there is no submitted form result, stop; never
continue with inferred answers. add_custom_carrier is only a routing
intent; carrier name, ID, SKU, revision, and custom flash config remain
downstream-owned.
Print the full indexed platform list before the questionnaire.
Explicit choices are only the first row from each chip-family group,
labeled with full-list index and product name, for example
Index 1 — Jetson AGX Orin 64GB.
For all other platforms, the user chooses Type something and enters
the numeric index from the printed list. Type something must be the
tool's built-in freeform row, not a custom explicit choice.
Prompt text: "Pick a platform shortcut, or choose Type something and
enter the numeric index from the full platform list printed above. Type
skip to skip platform selection and let downstream setup ask later."
Only platform shortcuts may be explicit choices. Do not add skip,
Other, or custom manual-entry choices; skip must be typed through
the built-in Type something path. Do not paraphrase Type something
as Other in prompt text.
Do not resolve flash-config variants here; pass the selected catalogue
row to jetson-init-target, which owns validation and flash-config
selection.
Each non-skip answer must be explicit and directly consumable by the
owning downstream skill; use skip for anything unknown. A BSP release
answer must be a concrete version token with an optional leading R and
two or three numeric components (for example R38.4, 38.2, 38.2.1,
or 36.4.4). Family placeholders such as R38.x or "pin a version
later" are not acceptable; use skip instead. Preserve the user's/source
token exactly when passing quick_start_prefill.download.bsp_release.
Image paths, source paths, document paths, carrier details, repository
overrides, toolchain choices, and document bindings are non-core setup
details. Leave them to the owning downstream skill. Downstream setup
skills should use documented defaults without asking when safe, and ask
only on conflicts, overwrites, incompatible releases, missing artifacts,
or user-requested overrides.
For Auto Setup, /jetson-download-bsp owns listing supported BSP releases
for the selected platform and warning on platform/BSP incompatibility.
Quick-start may forward a requested global candidate or typed release, but
it does not validate support or override compatibility decisions itself.
Show full SKILL.md (756 more words)Show less
Build Prefill Bundle
Normalize only explicit non-skip answers and keep them in memory:
yaml
quick_start_prefill:
mode: Auto Setup | Guided Setup | Use Existing Workspace
target: { active_platform, custom_carrier }
download: { bsp_release }
Omit any skipped field and omit owner subsets with no core answer. Validate
only enough to avoid obvious misrouting, such as blank mode or a BSP
placeholder. Downstream defaults are skill behavior, not prefill fields.
Do not invent missing identity values and do not write profile blocks from
quick-start.
Invoke Downstream Skills
Build the downstream plan from quick_start_prefill.mode, target state,
and prerequisite checks. Before invoking downstream skills, print a
non-blocking execution plan summary with the selected mode, target
platform intent, requested BSP release or downstream release-selection
handoff, custom-carrier intent, planned downstream skills in order, and
non-core details that downstream skills may still ask for. Do not ask to
approve this summary; continue dispatching unless the user chose cancel
in the core questionnaire. This does not bypass downstream blocking
validation gates.
Pass each downstream skill its relevant quick_start_prefill subset plus
the top-level mode.
The downstream skill must use valid prefilled answers, ask again for
missing, invalid, ambiguous, or incompatible required inputs, and own any
mutation of its profile block: reference_devkit:, custom_carrier:,
bsp_image:, source:, or documents:.
For Use Existing Workspace, verify:
<bsp_image.root_path or workspace/Image>/Linux_for_Tegra/ exists.
Linux_for_Tegra/rootfs/etc/nv_tegra_release exists, proving
apply_binaries.sh ran; otherwise route to /jetson-init-image.
<source.root_path or workspace/Source>/Linux_for_Tegra/ exists and is
a git repo.
recorded documents: paths exist when present.
Finish with executed skills, skipped skills, remaining downstream questions,
and whether the workspace is ready for customize-*.
Suggest I/O customization next steps
After the dispatch summary, print a non-blocking "Next steps —
I/O customization" list only when the active profile's documents:
block has at least one carrier-board-related slot bound via
/jetson-link-docs. The qualifying slots are:
documents.carrier_board_spec
documents.carrier_schematic
documents.ref_devkit_pinmux_xls
documents.custom_carrier_schematic
documents.custom_carrier_pinmux_xls
If none of those are bound, skip this step — the user has not yet
provided the carrier-board files these skills depend on, and the right
next step is /jetson-link-docs, not a customize-* skill. Say so in
one line ("No carrier-board docs bound — run /jetson-link-docs to
register the carrier schematic / pinmux xlsx before I/O customization")
and stop.
When at least one qualifying slot is bound, print the list below.
Adapting a board to its on-board peripherals typically starts with
pinmux and UPHY, then branches per-controller:
/jetson-customize-pinmux — per-pin SFIO / direction / pull state
(consumes documents.custom_carrier_pinmux_xls or
documents.ref_devkit_pinmux_xls).
/jetson-customize-uphy — UPHY lane allocation via ODMDATA
(consumes documents.module_design_guide and
documents.custom_carrier_schematic / documents.carrier_schematic).
For each suggested skill, flag any required documents.* slot that is
not yet bound as "run /jetson-link-docs first to bind <field>" rather
than hiding the skill — the user may want to fix the binding and retry.
Do not invoke the customize-* skills here; quick-start ends at the
suggestion.
Purpose
Single entrypoint that gathers the four routing-critical answers
(mode, active_platform, bsp_release, custom_carrier) and
dispatches the right downstream Setup skills. Not a one-stop installer
— it never downloads, extracts, materializes, or binds anything on its
own.
Prerequisites
bsp-platforms-catalogue.md reachable (used to build the platform
list).
Submitted accept_disclaimer answer from the disclaimer
AskUserQuestions form.
AskUserQuestions UI available — quick-start refuses to continue
without submitted disclaimer and core questionnaire forms.
Network access only required for the Auto Setup branch
(/jetson-download-bsp is the actual fetcher).
Limitations
Asks at most four core questions; any non-core value (image paths,
source paths, document paths, carrier IDs, toolchain) is left to the
owning downstream skill.
Does not validate platform / BSP-release compatibility — that is
/jetson-download-bsp's job.
Does not write to target-platform/*.yaml; profile mutation stays
with /jetson-init-target, /jetson-init-image, etc.
Troubleshooting
"questionnaire cancelled" exit — the form must be submitted; rerun
/jetson-quick-start and complete the four questions.
"disclaimer not accepted" exit — rerun /jetson-quick-start, read
the disclaimer, and choose accept_disclaimer in the disclaimer form
when ready to continue.
Active profile not auto-selected — by design; even with one
matching profile, the user must pick it explicitly in the form.
Unknown bsp_release typed — pass it as-is; /jetson-download-bsp
validates the release against the official archive.
Gotchas
Ask the core questionnaire once. Downstream skills ask for non-core
values, blanks, invalid values, or ambiguous choices.
Keep quick_start_prefill in memory unless the user asks to save it.
Never construct NVIDIA artifact URLs here; /jetson-download-bsp follows
links from the selected Jetson Linux archive release.
platform_template.yaml has no User Guide / Release Notes slots. Store
those files or URLs in the final summary or KB, not as invented profile
fields.
Jetson Quick Start next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
Workflow for porting an existing Triton kernel in fla/ops/ to Gluon (triton.experimental.gluon) to gain explicit control over tensor layouts, shared memory, async data movement (cp.async / TMA), MMA…
Preflight checks and diagnosis for ten known failure modes of ML training on NVIDIA DGX Spark's GB10, spanning launch errors, memory, thermals, bandwidth and precision.
Shows how to launch distributed Megatron-LM training on a SLURM cluster: sbatch skeleton, torch.distributed.run setup, CUDA_DEVICE_MAX_CONNECTIONS rules and failure diagnosis.
Sets up and runs NVIDIA Cosmos Policy evaluations on the LIBERO and RoboCasa simulators, including headless GPU rendering and inference latency profiling.
A skill your agent uses when the user wants to deploy, run, debug, tear down, or call the REST API of the RTVI-CV 2D detection / tracking microservice.
Generates, validates, compares and explains HOLOLINK_def.svh macro files for the HSB IP, using bundled Python scripts and asking before it writes anything.
Orchestrates defect image generation for PCBA, metal surface and glass inspection with NVIDIA Cosmos AnomalyGen on OSMO, from cold-start Day 0 to real-photo Day 1 labeling.
Orchestrates video data augmentation and auto-labeling workflows on OSMO, from flow selection and preflight checks to submission, monitoring and output download.
Runs NVIDIA TAO Data Services KPI analysis on object detection results, comparing predictions to ground truth and writing per-class precision, recall and AP to a CSV.
Entry skill for Jetson / IGX BSP customization. An agent skill from NVIDIA/skills. Jetson Quick Start is an agent skill from NVIDIA/skills, published by the product's own GitHub organization. Entry skill for Jetson / IGX BSP customization.
When should I use Jetson Quick Start?
Jetson Quick Start fits situations like: tasks that involve GPU and accelerator computing.
How do I install Jetson Quick Start in Claude Code?
Run `npx skills add NVIDIA/skills --skill jetson-quick-start -a claude-code`. Or copy the skill folder (skills/jetson-quick-start in NVIDIA/skills) into .claude/skills/jetson-quick-start in your project. Claude Code loads it when a task matches its description.
How do I install Jetson Quick Start in Codex?
Run `npx skills add NVIDIA/skills --skill jetson-quick-start -a codex`. Or copy the skill folder (skills/jetson-quick-start in NVIDIA/skills) into .agents/skills/jetson-quick-start in your project. Codex loads it when a task matches its description.
Can I use Jetson Quick Start in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add NVIDIA/skills --skill jetson-quick-start -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/jetson-quick-start, .gemini/skills/jetson-quick-start, .github/skills/jetson-quick-start and .opencode/skills/jetson-quick-start in your project.
What does Jetson Quick Start need to run?
SKILL.md names no scripts, command-line tools or credentials: Jetson Quick Start is instructions for the agent only.
Does Jetson Quick Start access the network?
SKILL.md names 1 domain. As links in the text: developer.nvidia.com. This is read from the text; nothing was executed.
Is Jetson Quick Start safe to install?
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
What licence does Jetson Quick Start use?
Jetson Quick Start is published under the Apache-2.0 licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Jetson Quick Start use?
About 4.7k tokens (SKILL.md is roughly 19k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Jetson Quick Start?
Skills that share tags, products or a category with Jetson Quick Start: Fla Triton To Gluon (fla-org/flash-linear-attention, 5.8k stars), DGX Spark Memory and Thermal Ops (wshobson/agents, 40k stars), DGX Spark Training Gotchas (wshobson/agents, 40k stars) and Megatron-LM on SLURM (NVIDIA/Megatron-LM, 18k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Jetson Quick Start?
NVIDIA (a GitHub organization, an official publisher) maintains it in NVIDIA/skills, which has 3,534 GitHub stars. The repository holds 380 skills in this directory. The repository was last updated on October 7, 2026.
Source: NVIDIA/skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.