What does this skill do, and when should you use it?
This skill targets applications where NeMo Relay is present but runtime behavior is missing or incorrect. It directs the agent to identify the failing runtime layer before changing configuration. Coverage includes native artifact loading, scope stacks, event registration, subscribers and exporters, plugins, adaptive components, provider payload conversion, and OpenTelemetry or OpenInference export. It is useful for integration troubleshooting, but it is not a complete installation guide or an automated repair tool.
Guides checks for Python, Node.js, and Go native components; verifies active scopes and scope stacks; reviews subscriber, exporter, middleware, plugin, and adaptive configuration; and checks registration timing, concurrency boundaries, lifecycle calls, and provider-object conversion. It also provides troubleshooting actions for Rust builds, Python environment reconstruction, Node native-addon rebuilds, Go FFI loader paths, ATIF output, OpenTelemetry/OpenInference exporters, and independent plugin validation.
- A developer whose Python application cannot import the NeMo Relay native extension.
- A developer debugging a Node.js native addon that was not built or will not load.
- A developer whose execution occurs outside an active scope or crosses a thread, task, goroutine, or worker boundary incorrectly.
- A developer investigating missing subscriber events, empty ATIF output, or mixed data from concurrent agents.
- A developer whose plugin or adaptive configuration appears enabled but does not change runtime behavior.
How do you install this skill?
- Before running uv sync, rebuilding native extensions, changing dynamic-loader paths, or modifying exporter configuration, confirm scope, permissions, and rollback options.
- The matrix lacks concrete version requirements, command arguments, and failure branches; validate against the deployed NeMo Relay version and platform documentation rather than treating it as a verified fix.
- Non-fit scenarios, Chinese-language behavior, and mainland-China network reachability are not explicitly defined.
- Shell / CLI
- Local filesystem
RustPythonuvNode.jsGo
Install the individual skill from NVIDIA's collection with the default skills CLI:
npx skills add nvidia/skills --skill nemo-relay-debug-runtime-integration --yes
You can also browse the catalog with npx skills add nvidia/skills --list. The README does not specify additional installation dependencies or version requirements for this skill.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/nemo-relay-debug-runtime-integration" ~/.claude/skills/
rm -rf "$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:
- Debug a NeMo Relay integration that produces no lifecycle events, starting by identifying the failing runtime layer.
Place the installed skill where the agent can load it, then invoke it for a NeMo Relay runtime integration problem, for example: "Debug a NeMo Relay integration that produces no lifecycle events, starting by identifying the failing runtime layer." The application code, binding versions, and platform configuration must be supplied by the user; SKILL.md defines no dedicated command or automation script.
What are this skill's strengths and limitations?
- Uses a runtime-layer-first diagnostic approach across multiple language bindings and integration paths.
- Covers concrete checks for scope isolation, event registration, concurrency, plugins, adaptive behavior, and exporters.
- Distinguishes managed execute, manual lifecycle, typed-wrapper, and adaptive/plugin API layers.
- Provides no runnable diagnostic script or complete repair implementation.
- Provides no binding-version compatibility matrix or verified platform list.
- Several recommendations depend on the application's build system and runtime environment, so the user must perform and interpret the checks.
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 |
|---|---|---|---|---|
| NeMo Relay Runtime Integration Debugger this page ✓ NVIDIA · Official | 56 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| NeMo Relay Call Instrumentation ✓ NVIDIA · Official | 55 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| NeMo Relay Typed Wrappers & Codecs ✓ NVIDIA · Official | 55 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| NeMo Relay Context Isolation ✓ NVIDIA · Official | 55 · Use with care | ★ 3.6k | 13d ago | Apache-2.0 |
| NeMo Relay Plugin Builder ✓ NVIDIA · Official | 60 · Recommended | ★ 3.6k | 13d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill is scoped to NeMo Relay runtime-integration diagnosis and does not request credentials or broad access. The skill card explicitly prohibits secrets in prompts, logs, and output and recommends least-privilege credentials. However, uv sync, native rebuilds, loader-path changes, and exporter configuration can affect the environment or external systems; user confirmation, rollback, data-flow disclosure, and detailed safety boundaries are incomplete, so points are deducted.
The failure classes and troubleshooting matrix are internally coherent and cover Python, Node, Go, Rust, scopes, concurrency, plugins, and exporters, with several diagnostic checks. Concrete commands, version prerequisites, abnormal-input handling, and clear post-failure branches are missing. This is a static review without committed tests covering this skill's key paths, so the score remains below the static ceiling.
The audience and activation scenarios are clear, covering load failures, scopes, events, plugins, adaptive behavior, and API-layer selection. Non-fit boundaries, input/output contracts, exclusion triggers, and Chinese-language support are not specified. Mainland-China reachability is not evidenced, although the core function is not stated to depend entirely on overseas services, so points are deducted.
The SKILL.md is concise and layered, with metadata, first checks, failure classes, a troubleshooting matrix, and related skills; the skill card adds license, dependency, and risk notes. The static report explicitly identifies missing Instructions and Examples sections. Version policy, changelog, maintenance ownership, and update path are unclear, and the author format also fails the reported schema check, so points are deducted.
The matrix directly addresses the three positive evaluation cases: Python native-extension import failure, missing Node lifecycle events, and Go cross-request scope leakage. The committed evaluation report shows strong results on the positive tasks and includes a negative-routing task. However, the guidance is mostly directional, lacks copy-ready commands, platform/version branches, and complete validation outputs; the report cannot substitute for static reproduction, so the score is conservatively capped and reduced.
SKILL.md, the evaluation report, and evals.json provide auditable failure mappings, expected behaviors, and limited evaluation evidence, while the report discloses static-validation findings. The evaluation has only four tasks with one attempt each and provides no independently reproducible logs, committed tests for this skill, or cross-source corroboration; confidence therefore remains low and the score stays below the static ceiling.
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 →