What does this skill do, and when should you use it?
This skill from NVIDIA’s skills repository handles DOCA version detection, four-way consistency checks, NGC container tag semantics, and version-related diagnosis. It guides an agent through comparing the host installation, application version, DOCA capability tool, and BlueField release information. It also covers capability availability by installed release and apt-source consistency before package installation. The skill is a thin loader; detailed procedures are delegated to the companion CAPABILITIES.md and TASKS.md files.
Directs the agent to read CAPABILITIES.md and TASKS.md; run or request results from pkg-config --modversion doca-common, /opt/mellanox/doca/applications/VERSION, doca_caps --version, bfver, and /etc/mlnx-release; validate the four-way version match; inspect NGC version semantics, version-matrix lookups, DOCA Compatibility Policy routing, and apt-source consistency; and diagnose partial installs, host–BFB mismatches, NGC mixing, and build-time/runtime drift.
- A DOCA developer needs to identify the version actually installed because the documentation and host appear inconsistent.
- A BlueField user needs to check whether the host DOCA release and BFB release are compatible.
- An engineer sees a successful build followed by DOCA_ERROR_NOT_SUPPORTED, an undefined symbol, no dataplane effect, or an unchanged counter.
- A container user needs to understand an NGC tag or determine whether host and container versions are being mixed.
- A maintainer is adding a DOCA library and needs a version-compatibility section or an updated version pin.
- A system administrator is planning a DOCA upgrade, downgrade, or package installation and must first verify the configured apt source.
How do you install this skill?
- Do not treat versions, package names, or tool paths in the skill text as facts about the target host; obtain and inspect the detection-chain output first.
- Before any apt install or upgrade, confirm repository sources, candidate versions, privileges, user consent, and a recoverable plan; these controls are not fully specified by the skill.
- The benchmark states that its source dataset was unavailable, and this review executed nothing; its results should not be treated as independent runtime validation.
- Key references and some capabilities depend on NVIDIA-hosted documentation, repositories, or NGC; verify mainland-China reachability and provide an offline alternative where needed.
- Shell / CLI
- Local filesystem
DOCA installation at /opt/mellanox/doca for validation workflowspkg-configdoca_capsbfver
Use the repository README’s skills CLI flow: npx skills add nvidia/skills --skill doca-version --yes. The README says that cloning the repository or manually copying skill directories is unnecessary; the skill becomes available when the agent encounters a relevant task.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/doca-version" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- What DOCA version do I actually have?
- pkg-config reports 3.3.0 but doca_caps reports 3.2.0—are they consistent?
- My program built but does nothing on the wire.
After installation, ask a concrete version-handling question such as “What DOCA version do I actually have?”, “pkg-config reports 3.3.0 but doca_caps reports 3.2.0—are they consistent?”, or “My program built but does nothing on the wire.” The skill routes procedural work to TASKS.md and version rules and safety policy to CAPABILITIES.md. It must not infer a version string from memory; validation requires a live DOCA installation at /opt/mellanox/doca.
What are this skill's strengths and limitations?
- Provides a unified check across host, application, capability-tool, and BlueField release information.
- Defines clear boundaries from installation, library-specific API, and general debugging skills.
- Addresses practical cases including build/runtime drift, NGC tags, and apt-source consistency.
- The README states that published skills are governed through NVIDIA signatures, evaluation datasets, and an automated sync pipeline.
- SKILL.md is a thin loader, so detailed procedures depend on CAPABILITIES.md and TASKS.md, whose contents are not included here.
- Validation workflows require a live DOCA installation under /opt/mellanox/doca.
- It does not install or select DOCA packages and does not replace library-specific API or general debugging skills.
- The supplied source does not show concrete evaluation results or test-case contents for this individual skill.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| DOCA Version Consistency Assistant this page ✓ NVIDIA · Official | 53 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| DOCA Programming Guide ✓ NVIDIA · Official | 54 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| DOCA Environment Setup ✓ NVIDIA · Official | 57 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| DOCA Telemetry Exporter Development ✓ NVIDIA · Official | 55 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| DOCA DPA High-Level Tracer ✓ NVIDIA · Official | 50 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill emphasizes source-based version detection, refusing to invent versions, separating build/runtime evidence, and routing installation or firmware actions elsewhere; however, it lacks systematic user confirmation, privilege boundaries, rollback procedures, and sensitive-data guidance, while including potentially state-changing apt recommendations, so points are deducted.
The detection chain, four-way match rule, error taxonomy, and escalation paths are generally coherent; however, this review is static, key commands were not executed, and behavior across DOCA releases, distributions, and installation layouts remains environment-dependent, so the static ceiling applies.
Audience, trigger phrases, scope boundaries, and routing rules are comparatively clear across BlueField, host, and NGC scenarios; Chinese-language support, mainland-China reachability, and nonstandard installations are not adequately addressed, and many workflows depend on NVIDIA documentation, repositories, or a live target host.
The loader, capabilities, task workflows, evaluation report, and skill card provide layered documentation, dependency notes, limitations, and ownership signals; however, SKILL.md lacks an author and recommended Instructions/Examples sections, license/version metadata is inconsistent across files and repository metadata, and the update path is not fully explicit.
The skill can directly guide version detection, build/runtime drift diagnosis, and routing through concrete commands and decision branches; nevertheless, results depend on user-side execution and companion files, the version-matrix data is incomplete, and static review cannot verify actual outputs, so the score is capped.
Committed evaluation records, assertions, command-level procedures, and NVIDIA documentation references provide some auditability; however, the evaluation dataset is unavailable, no execution was reproduced in this review, and independent CI or test-suite evidence is limited, so a higher score is not justified.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →