复制安装命令
用 Codex 或 Claude 安装复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它先审查 Skill 页面再帮你安装。
复制前请先查看来源、License 和安全提示。
Official, NVIDIA-verified Agent Skills for Claude Code, Codex, and other coding agents.
用 Codex 或 Claude 安装复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它先审查 Skill 页面再帮你安装。
复制前请先查看来源、License 和安全提示。
来源文件:README.md
Official, NVIDIA-verified Agent Skills for Claude Code, Codex, and other coding agents.
📖 Docs: docs.nvidia.com/skills · 📺 Livestream: From Vulnerable to Verified · 📝 Blog: NVIDIA Verified Agent Skills: Capability Governance for AI Agents
Skills are portable instruction sets that teach AI agents how to use NVIDIA software optimally: Physical AI and robotics workflows, simulation, CUDA-X libraries, RAG and AI Blueprints, and platform tools. This repository is a catalog: skills are maintained in their respective product repos, and mirrored here daily via an automated sync pipeline. Skills are being added continuously, so check back for updates. We are building this infrastructure in the open, and contributions are welcome. See the Roadmap for what is planned next.
Install NVIDIA skills with the default skills CLI flow:
npx skills add nvidia/skills
The CLI runs through npx and prompts you to choose a skill and install destination. You do not need to clone this repo or copy skill folders by hand.
Requires a current
skillsCLI (v1.5.16 or newer). Installing vianpx skills@latest add nvidia/skillsalways uses the latest. On older CLIs (v1.5.15 and earlier), skills may install but not appear in Claude Code — see Troubleshooting.
The skill is available the next time your agent loads skills and encounters a relevant task. For example, ask your agent to "solve a linear programming problem with cuOpt" and the skill guides it through the cuOpt Python API. In Claude Code, run /reload-skills to load newly installed skills in your current session.
Use this when you already know the skill name and want to skip prompts.
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --yes
Replace cuopt-numerical-optimization-api with any skill name from the Skill Catalog.
Use --agent to target a specific AI coding agent. Initially, we'll support common client targets, expanding the list over time. For the full list of clients supported by the spec, see the skills CLI Supported Agents table.
Claude Code
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent claude-code
Codex
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent codex
Snowflake CoCo
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent cortex
Cursor
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent cursor
Kiro
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent kiro-cli
Use --agent more than once to install the same skill into multiple agents.
npx skills add nvidia/skills \
--skill cuopt-numerical-optimization-api \
--agent claude-code \
--agent codex \
--agent cursor \
--agent kiro-cli
New skills land continuously, and existing ones are revised, renamed, or consolidated as the catalog evolves. Refresh what you have installed with:
npx skills update
Run it interactively and the CLI also flags skills that were removed or merged upstream (for example, when several skills are consolidated into one) and offers to remove the stale local copies. Use npx skills list to see what is installed and npx skills check to preview what is out of date first.
Use this when you want to see available NVIDIA skills before installing anything.
npx skills add nvidia/skills --list
For non-interactive installs, global installs, agent-specific installs, updates, removals, and fallback manual copying, see Advanced installation.
Where to file an issue depends on what's broken:
Per-product source repo links:
For issues with this catalog repo itself (README, structure, listing a new product): open an issue here.
Every published skill ships with a detached OMS signature (skill.oms.sig). The sync pipeline drops any skill missing the required artifacts before publishing, so every skill in the catalog carries:
SKILL.md — the skill instructions consumed by the agentskill-card.md — skill identity and governance cardskill.oms.sig — detached OMS signature (verifiable against nv-agent-root-cert.pem)evals/evals.json, evals/*.json, eval/*.json, or benchmark/evals.jsonBENCHMARK.md — generated benchmark report capturing verifiable uplift dataVerify a skill against the NVIDIA trust anchor nv-agent-root-cert.pem:
pip install model-signing
model_signing verify certificate SKILL_DIR \
--signature SKILL_DIR/skill.oms.sig \
--certificate_chain nv-agent-root-cert.pem \
--ignore_unsigned_files
A successful verification confirms that the skill contents have not been modified since signing by NVIDIA.
See Verify Signed Agent Skills for signature layout, the trust pipeline, and policy options.
NVIDIA/skills/
├── skills/ # NVIDIA-verified skills (count grows continuously),
│ │ synced from upstream product repos
│ ├── README.md # Browser-facing install guidance
│ ├── <product-prefix>-*/ # Flat layout — one dir per skill, product-prefixed
│ │ # e.g. aiq-*, cuopt-*, cupynumeric-*,
│ │ # dali-*, deepstream-*, dicom-*, digital-health-*,
│ │ # dynamo-*, earth2studio-*, holoscan-*, hsb-*,
│ │ # jetson-*, launch-nemo-rl, mcore-*,
│ │ # nemo-automodel-*, nemo-data-designer-plugin,
│ │ # nemo-evaluator-plugin, nemo-mbridge-* (20 skills),
│ │ # nemo-retriever, nemo-rl-* (4 skills),
│ │ # nemoclaw-user-guide, nemotron-*, nemotron-speech,
│ │ # nv-* (medical AI), physicsnemo-*, rag-*,
│ │ # skill-card-generator, tao-*, tilegym-*,
│ │ # vss-* (15 skills), accelerated-computing-cudf,
│ │ # cudaq-guide, portfolio-optimization
│ ├── omniverse-*/ # Physical AI — manually staged (see manual-components.yml)
│ └── physical-ai-*/ # Physical AI — manually staged
├── components.d/ # Product registry — one file per component, teams onboard here
│ ├── README.md # Schema and onboarding instructions
│ └── <product>.yml # one file per registered product
├── plugins/ # Packaged plugin distributions
│ └── nvidia-skills/ # Curated NVIDIA skills bundle (Claude Code, Codex)
├── plugins.d/ # Plugin build registry — config for `build-plugins.py`
│ ├── README.md
│ ├── _defaults.yml
│ └── nvidia-skills.yml
├── .claude-plugin/ # Claude Code marketplace metadata
│ └── marketplace.json
├── .agents/plugins/ # Agent marketplace metadata (other clients)
│ └── marketplace.json
├── docs/ # Long-form documentation (published via Fern)
│ ├── README.md # How to build the docs locally
│ ├── index.mdx
│ ├── advanced-install.mdx
│ ├── agent-skill-trust-pipeline.mdx
│ ├── release-checklist.mdx
│ ├── scanning-agent-skills.mdx
│ ├── signing-agent-skills.mdx
│ └── skill-cards.mdx
├── fern/ # Fern docs site configuration
├── .github/
│ ├── workflows/ # Sync pipeline, plugin validation, DCO check, author verify
│ └── scripts/ # regenerate-readme.sh, build-plugins.py,
│ # manual-components.yml (temp Physical AI catalog
│ # exception, removed after Computex 2026),
│ # marketplace/metadata.json (skill metadata sidecar)
├── nv-agent-root-cert.pem # Trust anchor for OMS signature verification
├── skills.sh.json # Skills.sh marketplace grouping config
├── CHANGELOG.md
├── CONTRIBUTING.md # Contribution guidelines
├── SECURITY.md # Security reporting policy
├── CODE_OF_CONDUCT.md # Community code of conduct
├── LICENSE-APACHE # Apache 2.0 (source code)
└── LICENSE-CC-BY-4.0 # CC BY 4.0 (documentation/skills)
Skills are maintained in their respective product repos (see the Source column in the Skill Catalog) and synced to this repo daily. Products only appear under skills/ after the sync pipeline confirms each skill carries:
skill.oms.sig — detached OMS-format signature (verifiable against nv-agent-root-cert.pem)skill-card.md — skill identity and governance cardevals/evals.json, evals/*.json, eval/*.json, or benchmark/evals.jsonWhen evaluation runs produce a BENCHMARK.md, it ships alongside the skill so consumers can see verifiable benchmark uplift data.
This repository adheres to the Agent Skills specification:
SKILL.md file at their root.name and description fields.skills-ref reference library.Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
This code is dual-licensed with documentation/skills under the CC-BY-4.0 AND source code under Apache-2.0 license terms. The full license texts can be found in LICENSE-APACHE and LICENSE-CC-BY-4.0 respectively.
license: Apache-2.0
name: doca-firefly
description: >
Use this skill when the user is operating the DOCA Firefly Service
container on BlueField — picking the four PTP configuration axes
(role / profile / domain / interface), wiring the BlueField PHC +
host follower + consumer workload pairing, deciding whether
PTP-grade time is even needed (vs. chrony / NTP), or debugging a
Firefly deployment where PTP isn't syncing or the host clock isn't
following. Trigger even when the user does not explicitly mention
"DOCA Firefly" or "PTP" — typical implicit phrasings include
"container green but PTP never advances past LISTENING", "Firefly
says synced but the host clock still drifts", "sync acquired but
offset is tens of microseconds", "my Rivermax SMPTE workload needs
PTP", or "is chrony good enough". Refuse and route elsewhere for
installing DOCA, host-side chrony / ptp4l config bodies, PTP
topology / boundary-clock design, building DOCA apps that read the
disciplined PHC, or other DOCA services (DMS, Flow-Inspector, HBN)
— those belong to other skills.
metadata:
kind: service
compatibility: >
BlueField-Arm-only DOCA service container; pulled from NVIDIA NGC
and started under the BlueField OS container runtime. Host-side
install is irrelevant. Requires a reachable PTP master (or runs as
the master itself) and a PTP-aware network path; the host-side
time follower (chrony / ptp4l / phc2sys reading the BlueField PHC)
is also operator-owned.Subsystem inventory (Run-12 correction, verified Run-13). DOCA Firefly is NOT just "a PTP daemon." The shipped
doca_firefly.yamlexposes six PTP-stack subsystems via environment variables, each with its own*_STATE,*_CONFIG_FILE, and (where relevant)*_INTERFACE/*_DEVICEknobs (the count is six because the PTP Monitor subsystem ships an internalphc2sysmonitor client that is distinct from the standalone PHC2SYS subsystem — both ship in the same container image):
- PTP (
PTP_STATE,PTP_INTERFACE,PTP_CONFIG_FILE) — theptp4ldaemon (or master, depending on profile) that drives the BlueField PHC.- PTP Monitor (
MONITOR_STATE,MONITOR_CONFIG_FILE,MONITOR_CLIENT_TYPE,MONITOR_CLIENT_PHC2SYS_INTERFACE,MONITOR_CLIENT_CONNECTION_TIMEOUT) — the monitor server + client surface; the internalphc2sysmonitor client (MONITOR_CLIENT_TYPE=phc2sys) is a real subsystem inside Firefly, not just a host-side concern.- PHC2SYS (
PHC2SYS_STATE,PHC2SYS_ARGS,PHC2SYS_CONFIG_FILE) — the container-internalphc2sysinstance; the bundle previously framedphc2sysas host-only, which is wrong.- PPS (
PPS_STATE,PPS_DEVICE) — the Pulse-Per-Second output (with the additionalenable_while_runninganddo_nothingstates beyond plain enable/disable).- SyncE (
SYNCE_STATE,SYNCE_INTERFACE,SYNCE_CONFIG_FILE) — Synchronous Ethernet frequency distribution; orthogonal to PTP.- Firefly Servo (
SERVO_STATE,SERVO_CONFIG_FILE) — the proprietary Firefly servo loop (alternative to the upstream linuxptp servo).The valid
PROFILEvalues are exactlydefault/media/telco-l2/custom(perdoca_firefly.yamlcomments) — the agent must not invent additional values. Subsystems configured asdefined_by_profileare controlled by the activePROFILE.Configuration-override env vars follow the pattern
CONF_<SUBSYSTEM>_<section>_<key>(e.g.CONF_PTP_global_priority1,CONF_SYNCE_global_backend,CONF_MONITOR_global_telemetry_export); these are the documented surface for overriding individual config keys without shipping a full custom config file.Configuration hierarchy: the mounted Firefly config file is mandatory and owns the primary PTP axes (role, profile, domain, interface, and transport).
CONF_<SUBSYSTEM>_<section>_<key>variables are optional, documented per-key overrides of that file; they are not a second standalone configuration model.
Where to start: This skill is for operating the DOCA Firefly
Service container, not for linking against a library. Firefly is
the PTP / PHC2SYS / PPS / SyncE / Servo / Monitor stack that
drives and observes the BlueField PTP Hardware Clock (PHC); it is
not the host-side time follower, not the consumer workload, and
not a programming surface. If the user wants to deploy the
container, open TASKS.md and start at
## configure. If the question is what shape
of service is Firefly and what PTP roles / profiles does it speak,
start at CAPABILITIES.md. If DOCA is not installed
on the BlueField yet, route to
doca-setup first. If the user's
real question is "I have a Rivermax SMPTE workload and the docs say I
need PTP", the right pairing is this skill plus
doca-rmax — Firefly disciplines
the PHC; Rivermax reads the disciplined time.
The CLASSES of Firefly questions this skill is built to answer, each with one worked example. The class is the load-bearing piece; the worked example is one instance.
CAPABILITIES.md ## Safety policy
TASKS.md ## configure.CAPABILITIES.md ## Capabilities and modes
TASKS.md ## configure.ptp4l /
Firefly says it's locked but chronyc tracking on the host shows
drift". Answered by the END-TO-END time-sync discipline in
CAPABILITIES.md ## Safety policy
TASKS.md ## configure.CAPABILITIES.md ## Safety policy
TASKS.md ## debug.CAPABILITIES.md ## Capabilities and modes
TASKS.md ## configure, which routes the
Rivermax side to
doca-rmax and refuses to
collapse the two services into one.SLAVE / MASTER state — was it role, domain, profile, or
interface?" — worked example: "container green but the
ports-state output never advances past LISTENING". Answered by
the four-axis-mismatch rule in
CAPABILITIES.md ## Error taxonomy
TASKS.md ## debug.This skill serves external operators and platform teams who deploy
the DOCA Firefly Service container to provide PTP-grade time
synchronization to time-sensitive workloads on BlueField + the host
behind it. Concretely: people running the Firefly container on
BlueField Arm, choosing its PTP role / profile / domain / interface
from the public Firefly guide, wiring the host-side follower (chrony
with the PHC source, or ptp4l reading the PHC) so the host clock
tracks the BlueField PHC, and validating the end-to-end discipline
before scaling a Rivermax, 5G UPF, financial-trading, or distributed-
database workload that depends on it.
It is not for NVIDIA developers contributing to Firefly itself,
and it is not a programming guide for building applications on
top of DOCA libraries (that is
doca-programming-guide
plus the matching libs/<library> skill). Firefly is a service,
not a library: the operator runs a container and configures PTP via
the documented config surface; they do not link against a
libfirefly.so to write their own program.
Path selection up front. Use Firefly when sub-microsecond, PTP-grade time precision is required on BlueField AND the host (SMPTE ST 2110 broadcast workloads layered on Rivermax, 5G UPF time requirements, distributed systems that need PTP-grade time, anything where NTP / chrony jitter is not tight enough). Do not reach for Firefly when NTP / chrony already meets the workload's time-precision budget, when no PTP-aware switching / boundary-clock infrastructure exists in the path, or when pure software-side time precision is sufficient — in those cases the correct answer is to keep the host's existing chrony / NTP setup and route the agent away from Firefly, not to deploy it speculatively.
Load this skill when the user is doing hands-on Firefly deployment work on a BlueField where DOCA is already installed. Concretely:
PROFILE env var accepts EXACTLY default / media /
telco-l2 / custom per services/firefly/doca_firefly.yaml;
these map onto industry PTP profile names: default → IEEE 1588,
media → SMPTE 2059-2, telco-l2 → G.8275.1 only (G.8275.2
corresponds to the separate telco-l3 config, reached via
custom) — do NOT put the industry names directly into the env
var), domain number,
network interface — for the user's deployment.ptp4l /
phc2sys reading the PHC) — without this step the host clock
does NOT follow the Firefly-disciplined PHC, regardless of how
cleanly Firefly comes up.Do not load this skill for general DOCA orientation, install of
DOCA itself, library-API questions, or non-PTP time topics. For
those, route via
doca-public-knowledge-map,
doca-setup, or the matching
libs/<library> skill.
This is a thin loader. Substantive material lives in two companion files:
CAPABILITIES.md — Firefly's architecture (container that drives
the BlueField PHC and speaks PTP on the wire), the four PTP
configuration axes (role / profile / domain / interface, with
transport as a fifth knob), the deployment shape (container on
BlueField Arm per the public Container Deployment Guide), the
pairing surface (Rivermax + host-side time-sync follower), the
observability surface (container logs + PHC offset + ports state),
the error taxonomy (four-axis-mismatch / host-follower / PTP-aware-
path / container-runtime), and the safety policy (PTP-vs-NTP path
selection, END-TO-END discipline, smoke-before-scale).TASKS.md — step-by-step workflows for the in-scope Firefly
verbs: configure, build, modify, run, test, debug,
plus a Deferred task verbs block routing out-of-scope questions
and a Command appendix of recurring commands.The skill assumes a BlueField where DOCA is already installed and
the operator has the privileges the public Firefly Service Guide
expects to pull, run, and configure containers on BlueField Arm.
It does not cover installing DOCA — that path goes through
doca-setup.
This skill is agent guidance, not a templates or sample-config bundle. To keep the boundary clean, it deliberately does not contain — and pull requests should not add:
doca-public-knowledge-map ## DOCA services;
Firefly's image tag is version-bound and changes between DOCA
releases. Inventing or memorizing a tag is the canonical
hallucination failure mode for a service skill.ptp4l / phc2sys config files.
Those are host-environment-specific and live on the host, not
inside the Firefly container. The skill names that the
host-side follower must be wired and what its source must be
(the BlueField PHC); the chrony / ptp4l config bodies belong
to the host operator and to upstream Linux PTP documentation.samples/, templates/, or reference/ subtree of any
kind. A mock or incomplete artifact in this skill's tree, even
one labeled "reference", is misleading: operators will read it
as production-ready.SKILL.md first to confirm the user's question is in
scope and that Firefly is the right answer at all (vs. keeping
NTP / chrony on the host).doca-public-knowledge-map
— the routing table to the public DOCA Firefly Service Guide and
the rest of the public DOCA documentation set. The Firefly URL is
listed under
## DOCA services.doca-setup — env preparation and
install verification on the BlueField where the Firefly container
will run, including the I have no install yet path via the
public NGC DOCA container. This skill assumes its preconditions
are satisfied on BlueField Arm.doca-version — canonical DOCA
version-handling rules. Firefly's container tag is version-bound;
this skill's ## Version compatibility cross-links the four-way
match rule and adds the container-tag-lags-host-package overlay.doca-structured-tools-contract —
the bundle's structured-tools precedence rule (detect / prefer /
fall back / report). The Command appendix in TASKS.md
honors this contract.doca-programming-guide
— general DOCA patterns. Firefly is service-shaped not library-
shaped, so the build / modify / first-app pattern there does not
apply directly, but the cross-library debug discipline (frontend-
before-backend, env-before-program) remains useful when Firefly
reports an error that originated in the container runtime or in
a DOCA library it called.doca-rmax — the canonical
paired workload. SMPTE ST 2110 Rivermax streams depend on a
Firefly-disciplined PHC; Firefly is the time-source side and
Rivermax is the timing-precise data-plane side. The two skills
load together for any broadcast-style deployment, and they do
NOT collapse into one another — Firefly does not stream media;
Rivermax does not discipline the PHC.doca-dms — sibling service skill. The
agent reading both skills should see the same service-skill shape
(container, BlueField Arm, deployment pattern, smoke-before-scale,
env preconditions, config schema) layered on top of a different
per-service domain (DMS = device management via gNMI / gNOI;
Firefly = time synchronization via PTP).doca-debug — the cross-cutting
debug ladder (install / version / build / link / runtime /
program / driver). Firefly-specific debug (PTP not syncing,
host clock not following, jitter past spec) overlays on top of
that ladder.
评论 (0)
暂无评论,成为第一个评论者吧!