dynamo-interconnect-check

Validate that a Dynamo deployment's NIXL/UCX/NCCL interconnect is ready for disaggregated serving over RDMA/NVLink. Use after recipe-runner brings a deployment up (especially disagg/multi-node) to confirm the KV transport is correct; use troubleshoot for diagnosing already-failed pods.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

The canonical home for this skill is dynamo-interconnect-check in NVIDIA/skills

Quality

Content

85%Weight 40%Scale 1-3

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, highly actionable read-only validation skill with a clear sequenced workflow, explicit output interpretation, and proper externalization of detail into a reference file. The main weakness is redundancy: the three subcommands are demonstrated three separate times.

Suggestions

Collapse the Available Scripts table and Examples section into the Instructions steps, or keep only one of them, to remove the triplication of the env/node/nixl commands and respect the token budget.

If retaining the Examples section, have it show only values not already given in Instructions (e.g. a multi-node namespace/pod pair) instead of repeating the same single-node examples.

Resolve or guard the BENCHMARK.md reference (the file is absent in this bundle); either remove the pointer or note it is auto-generated and may not be present locally.

Dimension	Reasoning	Score
Conciseness	The body is operational and free of concept-padding, but the same three subcommands (env/node/nixl) are repeated across the Instructions, the Available Scripts table, and the Examples section, which is tightening opportunity that keeps it below the lean-and-efficient anchor 3.	2 / 3
Actionability	Provides fully executable commands with concrete real values (e.g. recipes/qwen3-coder-480b/sglang/disagg, dynamo-demo, qwen-worker-0) plus run_script() protocol examples that are copy-paste ready, matching the fully-executable anchor.	3 / 3
Workflow Clarity	Three numbered steps are clearly sequenced, the Output Contract defines ok/warn/fail/skipped verdicts for interpretation, and the Troubleshooting table supplies symptom->cause->next-step feedback loops; this read-only skill needs no destructive-operation checkpoint cap.	3 / 3
Progressive Disclosure	SKILL.md is an overview that points to verified one-level-deep bundle files (references/interconnect-env-vars.md and scripts/check_interconnect.py) with clear signaling, and the env-var catalog is appropriately externalized rather than inlined.	3 / 3
	Total	11 / 12 Passed

Description

100%Weight 40%Scale 1-3

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, concrete description that states the action, the specific transport targets, and an explicit use-trigger, while cleanly distinguishing itself from the troubleshoot skill. Voice is appropriately third-person/imperative with no first- or second-person phrasing.

Dimension	Reasoning	Score
Specificity	Names two concrete actions ("Validate that a Dynamo deployment's NIXL/UCX/NCCL interconnect is ready", "confirm the KV transport is correct") against specific objects (NIXL/UCX/NCCL, RDMA/NVLink, KV transport), matching the multiple-specific-actions anchor rather than the partial anchor 2.	3 / 3
Completeness	Explicitly answers both what (validate interconnect readiness for disaggregated serving) and when ("Use after recipe-runner brings a deployment up (especially disagg/multi-node)"), matching the anchor-3 example pattern with an explicit trigger clause.	3 / 3
Trigger Term Quality	Uses the natural terms this audience actually says ("disagg", "multi-node", "recipe-runner", "RDMA/NVLink") with an explicit "Use after recipe-runner brings a deployment up" trigger, giving good coverage rather than just some relevant keywords.	3 / 3
Distinctiveness Conflict Risk	Occupies a clear niche (Dynamo disagg/multi-node fabric validation) and explicitly disambiguates from the sibling skill ("use troubleshoot for diagnosing already-failed pods"), so it is unlikely to trigger for the wrong skill.	3 / 3
	Total	12 / 12 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning

	Total	14 / 16 Passed

Repository: openai/plugins
Path: plugins/nvidia/skills/dynamo-interconnect-check/SKILL.md
Commit: 11c74d6

Reviewed: 17 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.