doca-upgrade
Changing DOCA versions safely: upgrading or downgrading the host DOCA packages, refreshing the BlueField BFB, bumping the NGC DOCA container tag, or rolling…
Skill by NVIDIA <doca@nvidia.com>
Chalk Labs
Need help deploying doca-upgrade?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/nvidia-doca-upgrade
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, frontend
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/doca-upgrade
What's inside
- Skill name
doca-upgrade- Description
- Changing DOCA versions safely: upgrading or downgrading the host DOCA packages, refreshing the BlueField BFB, bumping the NGC DOCA container tag, or rolling back, with the discipline detect -> report -> ASK -> only then a guided upgrade. Trigger on "is there a newer DOCA", "I want the newest DOCA features, can I just apt upgrade in place", "should I move to the next release", "the upgrade stopped midway and dpkg is in a broken state", "my component is being deprecated, what now", or "roll me back", even without the word upgrade. Detect what is installed, find the newer release, report the gap, then STOP and ask for explicit confirmation; never upgrade automatically. Do not use it for a runtime error on an unchanged install (doca-debug). Version detection alone belongs to doca-version, first-time install to doca-setup, any hardware / firmware / reboot step to doca-hardware-safety, and public-docs or sunset routing to doca-public-knowledge-map.
- License
- Apache-2.0
- Compatibility
- No DOCA install required to read this skill (it is a cross-cutting overlay loaded against any DOCA artifact skill); the detection, gap-report, and guided-upgrade steps within DO require a live DOCA install at /opt/mellanox/doca, and the hardware / firmware / reboot steps require a BlueField DPU or ConnectX NIC plus out-of-band console reachability as gated by doca-hardware-safety.
- Metadata
- kind: library · author: NVIDIA <doca@nvidia.com> · doca_version: 3.5
- Instructions
- 263 lines in SKILL.md · view on GitHub
Files (7)
- BENCHMARK.md
- CAPABILITIES.md
- SKILL.md
- TASKS.md
- evals.json · evals/
- skill-card.md
- skill.oms.sig
SKILL.md (first lines)
# DOCA upgrade ## Examples - **"Should I move to the next DOCA release?"** — detect what is installed, report the gap, then STOP and ask. Never auto-upgrade. - **"apt upgrade drifted my DOCA packages."** — this is recovery, not a silent repair. Detect, report, ask, then guide. ## Quick answers (answer from this file) - **Not changing versions** (a runtime error on the current, pinned install)? This skill does not apply: name `doca-debug` and stop. These two answers are complete here; open `TASKS.md` only for a guided upgrade after the user confirms. Run only the read-only detection commands yourself. Package commands (`sudo dpkg`, `apt`, `apt-get`) need root and change the system: give them to the user to run and stop; in an agent shell `sudo` waits for a password and the session stalls. - **"Can I just `apt upgrade` to the next DOCA?"** Never upgrade automatically. Detect the installed release with the `doca-version` four-source chain (`pkg-config --modversion doca-common`, `cat /opt/mellanox/doca/applications/VERSION`, `doca_caps --version`, and the BFB version on BlueField), report the gap to the target release, then STOP and ask for explicit confirmation. BFB, firmware or reboot steps follow `doca-hardware-safety`. - **"The upgrade aborted half-way and dpkg reports an interrupted transaction."** Recover the package database first: after the user explicitly confirms (both commands change system package state), the user runs `sudo dpkg --configure -a`, then `sudo apt-get -f install` (Debian/Ubuntu). Do not re-run the upgrade as the first action; re-detect the release with `doca-version` before deciding the next step. ## Instructions Follow Loading order below. Open the tasks companion for detect → report → ASK → guided upgrade. Open the capabilities companion for the upgrade-mode taxonomy and refuse-auto-upgrade rule. Do not open every companion file. **Where to start:** This skill is the bundle's single source of truth for the discipline that wraps a DOCA *upgrade or downgrade*. Open the tasks companion when the user wants to *do* something with an upgrade (detect the gap, ask for confirmation, walk the guided upgrade, verify it, or diagnose a failed one); open the capabilities companion when the question is *what does an upgrade even cover* (the host apt-upgrade vs BFB reflash vs container-tag bump vs downgrade/rollback modes, the never-auto rule, the confirmation gate, the sunset-awareness concern, the error taxonomy, and the upgrade-specific safety overlay). **How to use this skill.** Read this file, then search the companion file for the `## <name>` heading the question needs and read that whole section. Answer completely from it, quoting only its symbols and commands. Do not browse docs; use the entry points in `doca-public-knowledge-map`. Give environment checks as commands for the user's DOCA host. For out-of-scope requests, name the skill that owns them and stop. **For a planned move, the single most important rule this skill exists to enforce is detect → report → ASK → only-then guided upgrade. Never upgrade automatically.** The skill detects the installed DOCA env and what newer release is available, reports the gap clearly, then STOPS and asks the user for explicit confirmation. Only on an explicit "yes" does it walk a guided upgrade. A failed or partial move instead enters the recovery ladder without presenting a new target move. In both paths every hardware / firmware / reboot step is delegated to `doca-hardware-safety/SKILL.md`, never redefined here. Version detection itself is owned by `doca-version/SKILL.md`; this skill routes there and does not restate the detection chain. ## Example questions this skill answers well The CLASSES of upgrade questions this skill is built to answer, each with one worked example. The agent should treat the *class* as the load-bearing piece — the worked example is a single instance. - **"Is there a newer DOCA than what I have, and how far behind am I?"** — worked example: *"`pkg-config --modversion doca-common` says `<installed X.Y.Z>`; what's current and what does moving cost me?"*. Answered by the detect-and-report-the-gap workflow in ## configure, which routes the detection to `doca-version/SKILL.md` and the *what is current* lookup to `doca-public-knowledge-map/SKILL.md`. - **"Should I just run `apt upgrade` to move to the next DOCA release?"** — worked example: *"I want the newest Flow features; can I upgrade in place right now?"*. Answered by the never-auto rule and the confirmation gate in ## Safety policy plus the ask-before-acting step in ## run — the agent reports the gap and STOPS for explicit confirmation before any upgrade command runs. - **"My upgrade left the host on a different release than the BlueField — what happened?"** — worked example: *"after the upgrade `pkg-config` reports the new release but `bfver` still reports the old one"*. Answered by the host/BFB-skew row in ## Error taxonomy and the failed-upgrade diagnosis in ## debug, which re-routes the detection to `doca-version ## test`. - **"The upgrade aborted half-way and now `dpkg` is wedged — how do I recover?"** — worked example: *"`apt upgrade` was interrupted; `dpkg` reports an interrupted transaction"*. Answered by the partial-upgrade and aborted-transaction rows in ## Error taxonomy plus the recovery ladder in ## debug. - **"The upgrade needs me to reflash the BFB / flip BlueField mode / reboot — is that safe?"** — worked example: *"the target release requires a new BFB; can I reflash now?"*. Answered by routing every hardware / firmware / reboot step to `doca-hardware-safety/SKILL.md` via ## run; this skill never redefines the reflash, mode-flip, or cold-power-cycle discipline. - **"My installed component is being sunset — should I keep building on it or move?"** — worked example: *"I rely on DOCA App Shield / the DOCA Flow Inspector service; are they on a
From NVIDIA/skills, Apache-2.0.
More from NVIDIA/skills
- 0.846score
nemo-fabric-integrate
NVIDIA Corporation and Affiliates
Use this skill when integrating NVIDIA NeMo Fabric into a consumer application, service, evaluation harness, or platform through the typed Python SDK —…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
jetson-video-pipeline
Vinit Bansal <vinitkumarb@nvidia.com>
Use when planning, executing, and independently validating Jetson Video Codec SDK or PyNvVideoCodec encode/decode, transcode, segmentation, container decode…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nvflare-convert-pytorch
NVIDIA FLARE Team <federatedlearning@nvidia.com>
Convert existing plain or manual PyTorch training code into an NVFLARE federated job using Client API model exchange, local validation, and job export; use…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nemo-relay-get-started
NVIDIA Corporation and Affiliates
Use this skill when first-time NeMo Relay users want help selecting and following an appropriate quick-start guide or verifying their first working integration.
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
jetson-video-recipe
Vinit Bansal <vinitkumarb@nvidia.com>
Use when turning a Jetson encoder use case into one surface-neutral recipe with native Video Codec SDK and PyNvVideoCodec projections.
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nvidia-broadcast
NVIDIA Broadcast Team <broadcast-skills-dev@nvidia.com>
Use when controlling NVIDIA Broadcast through MCP to apply effects, process local media, select devices, or change camera resolution, or when Broadcast is…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
jetson-video-setup
Vinit Bansal <vinitkumarb@nvidia.com>
Use when installing, repairing, reusing, inspecting, or verifying readiness of the native NVIDIA Video Codec SDK or PyNvVideoCodec on Jetson, including the…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nvflare-diagnose-job
NVIDIA FLARE Team <federatedlearning@nvidia.com>
Use when the user asks why a reported NVFLARE job failure signal occurred: the job failed, stalled, timed out, lost clients, ended with EXECUTION_EXCEPTION…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nvflare-convert-lightning
NVIDIA FLARE Team <federatedlearning@nvidia.com>
Convert existing PyTorch Lightning training code into an NVFLARE federated job using the Lightning Client API patch, local validation, and job export; use…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nemo-relay-install
NVIDIA Corporation and Affiliates
Use this skill when choosing or running NeMo Relay installation for an agent runtime, application language, framework integration, or plugin bundle, or…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
nvflare-fed-stats
NVIDIA FLARE Team <federatedlearning@nvidia.com>
Compute federated statistics over tabular data (count, sum, mean, stddev, var, histogram, quantile, noise-protected min/max) and image data (count…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.846score
jetson-video-capability
Vinit Bansal <vinitkumarb@nvidia.com>
Use when Jetson codec, profile, chroma, bit-depth, dimension, engine-count, or operational support must be reconciled from live APIs, authenticated NVIDIA…
SkillAFreshness: A★ 3.7KClaude Code +9
Signals
- Stars
- 3,718
- Forks
- 457
- Open issues
- 8
- Last commit (this folder)
- Oct 22, 2026 (8 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/NVIDIA/skills
- Website
- docs.nvidia.com/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from NVIDIA <doca@nvidia.com>
- 0.790score
doca-bf4-deployment
NVIDIA <doca@nvidia.com>
WARNING: guides IRREVERSIBLE BlueField-4 hardware operations (PLDM firmware burns, ISO reflashes, power cycles, BMC factory resets); every mutating step is…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.790score
doca-container-deployment
NVIDIA <doca@nvidia.com>
Running an in-bundle DOCA service as a container on a BlueField (Argus, DMS, Firefly, or UROM service): kubelet standalone watching a static-pod manifests…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.790score
doca-bf3-deployment
NVIDIA <doca@nvidia.com>
Use this skill for BlueField-3 (BF3) day-1 platform bring-up via the classic RShim/BFB path: pushing a BlueField bundle (BFB) to the DPU over RShim with…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.790score
doca-bare-metal-deployment
NVIDIA <doca@nvidia.com>
Running a DOCA-linked binary directly on the hardware, with no container or kubelet: on host x86 with a BlueField NIC over PCIe, or bare-metal on BlueField Arm.
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.790score
doca-collectx-deployment
NVIDIA <doca@nvidia.com>
Use this skill to deploy and operate a CollectX (clx) based DOCA telemetry collector on a host or BlueField — wiring providers / counters into the collector…
SkillAFreshness: A★ 3.7KClaude Code +9 - 0.790score
doca-verbs
NVIDIA <doca@nvidia.com>
Raw verbs inside DOCA Core (doca-verbs): direct control of QP, CQ, PD, SRQ and address-handle objects through doca_verbs_context and doca_verbs_qp_create /…
SkillAFreshness: A★ 3.7KClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get doca-upgrade running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
doca-upgrade skill. Agent Mart. https://agentmart.dev/skills/nvidia-doca-upgrade (last updated 2026-10-24)