deployment

Serve a quantized or unquantized LLM checkpoint as an OpenAI-compatible API endpoint using vLLM, SGLang, or TRT-LLM. Use when user says "deploy model", "serve model", "start vLLM server", "launch SGLang", "TRT-LLM deploy", "AutoDeploy", "benchmark throughput", "serve checkpoint", or needs an inference endpoint from a HuggingFace or ModelOpt-quantized checkpoint. Do NOT use for quantizing models (use ptq) or evaluating accuracy (use evaluation).

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

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.

The body is highly actionable and well-structured with a clear decision flow and validation checkpoints, and it practices progressive disclosure with real one-level-deep references. The main weakness is conciseness: two long inline framework-cookbook notes could be moved to their reference files.

Suggestions

Move the sm_103 NVFP4 advisory (lines 128-138) and the SGLang cookbook paragraph (lines 152-160) into references/vllm.md and references/sglang.md respectively, leaving a one-line pointer in SKILL.md to recover the conciseness-3 anchor.

The error-handling table duplicates some guidance already in references/support-matrix.md and references/unsupported-models.md; consider trimming to the rows not covered there.

Dimension	Reasoning	Score
Conciseness	Mostly lean and command-driven, but large inline advisory blocks (the sm_103 NVFP4 blockquote and SGLang cookbook paragraph, lines 128-160) teach framework internals Claude largely already knows and could be moved to references; below the 'every token earns its place' anchor.	2 / 3
Actionability	Provides executable quick-start commands (deploy.sh invocations, vllm/sglang server commands), a TRT-LLM Python snippet, and concrete verify curl calls — copy-paste ready as the anchor requires.	3 / 3
Workflow Clarity	Clear numbered decision flow (0-6) with explicit validation checkpoints (Step 5 verify-all-checks-must-pass, container-auth gate before SLURM submit) and a feedback-loop style error-handling table, matching the 'clear sequence with explicit validation' anchor.	3 / 3
Progressive Disclosure	SKILL.md is an overview that signals one-level-deep references (references/vllm.md, sglang.md, trtllm.md, support-matrix.md, benchmarking.md, etc.) — all verified to exist — with a primary script and common-skill pointers, matching the 'clear overview with well-signaled one-level-deep references' anchor.	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.

The description is specific, trigger-rich, and explicitly scoped, with a clear what/when structure and positive deconfliction against sibling skills. Voice is appropriately third person.

Dimension	Reasoning	Score
Specificity	Lists concrete actions ('Serve ... checkpoint as an OpenAI-compatible API endpoint') with specific frameworks (vLLM, SGLang, TRT-LLM) and explicit source types (HuggingFace, ModelOpt-quantized), matching the 'multiple specific concrete actions' anchor.	3 / 3
Completeness	Explicitly answers what (serve checkpoint as OpenAI-compatible endpoint) and when (the 'Use when user says...' clause), plus an explicit 'Do NOT use for...' negative-scope clause, satisfying the 'clearly answers both' anchor.	3 / 3
Trigger Term Quality	Dense natural trigger coverage ('deploy model', 'serve model', 'start vLLM server', 'launch SGLang', 'TRT-LLM deploy', 'AutoDeploy', 'benchmark throughput', 'serve checkpoint') — terms a user would naturally say, meeting the 'good coverage' anchor.	3 / 3
Distinctiveness Conflict Risk	Distinct niche (inference serving of quantized checkpoints) plus explicit disambiguation against ptq and evaluation skills ('Do NOT use for quantizing ... or evaluating accuracy') makes wrong-skill conflict unlikely.	3 / 3
	Total	12 / 12 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: NVIDIA/Model-Optimizer
Path: .agents/skills/deployment/SKILL.md
Commit: 33d05b0

Reviewed: 15 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.