vllm-optimize

vllm-optimize is a skill for Claude Code, Codex from AMD-AIM/inference-skill. It costs 49 tokens per session (530 once invoked), scanned A, original, MIT.

A workflow for measuring and investigating vLLM, a system for serving AI models, in containers. It runs model-serving benchmarks and examines how the GPU spends its time.

In plain words
What is it for?
Use it to run benchmark sweeps, collect profiling traces, and analyze GPU kernel performance on AMD or NVIDIA hardware.
Why use it?
It helps show how response speed changes with concurrent requests and where GPU execution is slowing the service down.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/amd-aim/inference-skill/vllm-optimize
Any agent
npx skills add AMD-AIM/inference-skill --skill vllm-optimize
Clone the repo
git clone --depth 1 https://github.com/AMD-AIM/inference-skill

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for vllm-optimize

README.md
[![agentmods](https://agentmods.dev/badge/skills/amd-aim/inference-skill/vllm-optimize.svg)](https://agentmods.dev/skills/amd-aim/inference-skill/vllm-optimize)
Your own site
<a href="https://agentmods.dev/skills/amd-aim/inference-skill/vllm-optimize"><img src="https://agentmods.dev/badge/skills/amd-aim/inference-skill/vllm-optimize.svg" alt="Measured on agentmods" height="20"></a>
Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 530 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00049 $0.00530
Opus 5 $0.00024 $0.00265
Sonnet 5 $0.00010 $0.00106
Haiku 4.5 $0.00005 $0.00053

Measured 3d ago against content hash c3d28b29ede6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

vllm-optimize scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/vllm-optimize/SKILL.md · 70 lines

What it actually says

vLLM Optimize

Automated vLLM inference benchmark and profiling workflow for containerized environments.

Quick Start

Run the workflow with a model name:

use vllm-optimize skill for Qwen/Qwen3.5-35B-A3B

The skill will automatically:

  1. Start vLLM server with the specified model
  2. Run benchmark at various concurrency levels
  3. Generate profiling traces
  4. Analyze GPU kernel performance

First-turn Latency Rule

  • Do not read any other file before the first visible reply
  • Send one short kickoff status update explaining the workflow
  • Ask the first grouped setup form with options

Guided Setup Flow

  1. Start with one short high-level question round:

    • Run plan (smoke test vs full sweep)
    • Output (where to save results)
    • GPUs (which GPUs to use)
  2. After Round 1 answers, read INTAKE.md for deeper config

  3. Read RUNTIME.md for execution bootstrap

  4. Summarize the final plan and get confirmation before executing

  5. After confirmation, start execution following phase docs

Modes

  • full: benchmark + profiling + analysis
  • benchmark: benchmark only (faster)
  • profile: profiling only (requires server running)

Files to Read

  1. Before Round 1: no extra file reads required
  2. After Round 1 answers: INTAKE.md
  3. Before execution: RUNTIME.md
  4. Phase docs: phases/*.md

References

Files

What ships with it

8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 70 lines · 49 tokens per session scan A c3d28b29ede6

Subscribe to this mod's changes

vllm-optimize is a skill published in the GitHub repository AMD-AIM/inference-skill (5 stars, last pushed 3mo ago), licensed MIT. It adds 49 tokens to every session and 530 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

instrument-data-to-allotrope

Convert laboratory instrument output files (PDF, CSV, Excel, TXT) to Allotrope Simple Model (ASM) JSON format or flattened 2D CSV. Use this skill when scientists need to standardize instrument data for LIMS systems, data lakes, or downstream analysis. Supports auto-detection of instrument types. Outputs include full…

anthropics/knowledge-work-plugins · 123 tokens

mixed-precision

Use FP16/BF16 mixed precision to accelerate training and reduce memory. Use when optimizing GPU performance.

aiming-lab/AutoResearchClaw · 25 tokens

oracle

Best practices for using the oracle CLI (prompt + file bundling, engines, sessions, and file attachment patterns).

the-open-agent/openagent · 25 tokens

summarize-document

Summarize a long PDF, scan, office file, text file, or markdown file through the connected Superlinked MCP edge instead of reading the whole source into model context. Use when the user asks for a summary, overview, digest, or "what does this document say" about a large file.

superlinked/sie · 66 tokens

generate-codebook

Generate a citable data dictionary / codebook from a tabular dataset (CSV/TSV/Excel/Parquet/Stata/SAS). Profiles every variable — role, type, units placeholder, level frequencies, range/quantiles, missingness — and emits codebook.md + codebook.json. Flags coded variables whose level meanings are unknown as [NEEDS…

Aperivue/medsci-skills · 97 tokens

nvalchemi-model-wrapping

How to wrap an arbitrary MLIP (Machine Learning Interatomic Potential) using the BaseModelMixin interface to standardize inputs, outputs, and embeddings. Use when integrating a model such as MACE or AIMNet2 (e.g. MACEWrapper, loading pretrained checkpoints) so dynamics, training, or fine-tuning stages can call it, or…

NVIDIA/nvalchemi-toolkit · 94 tokens