What is infx-bench-script-generator?

A script generator for running an InferenceX benchmark recipe manually on a local GPU server. It creates scripts for starting a Docker container, starting the server, and replaying an AgentX client, without Slurm or srt-slurm.

It reads a YAML recipe and override, checks that the override exists, and generates scripts with optional model, repository, trace, result, and profiler paths.

It removes the need to recreate a scheduled-cluster benchmark setup by hand when testing locally.

infx-bench-script-generator — Skill by shen-shanshan/vllm-dev-skills

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.

Any agent

npx skills add shen-shanshan/vllm-dev-skills --skill infx-bench-script-generator

Clone the repo

git clone --depth 1 https://github.com/shen-shanshan/vllm-dev-skills

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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/shen-shanshan/vllm-dev-skills/infx-bench-script-generator/github.svg)](https://agentmods.dev/skills/shen-shanshan/vllm-dev-skills/infx-bench-script-generator)

Your own site

<a href="https://agentmods.dev/skills/shen-shanshan/vllm-dev-skills/infx-bench-script-generator"><img src="https://agentmods.dev/badge/skills/shen-shanshan/vllm-dev-skills/infx-bench-script-generator/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/shen-shanshan/vllm-dev-skills/infx-bench-script-generator"><img src="https://agentmods.dev/badge/skills/shen-shanshan/vllm-dev-skills/infx-bench-script-generator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,516 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.00000	$0.01516
Opus 5.5	$0.00000	$0.00606
Sonnet 5.5	$0.00000	$0.00303
Haiku 4.5	$0.00000	$0.00152

Measured 27d ago against content hash 0f7b7fa8cfce, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

infx-bench-script-generator 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 27d ago.

The scan reads SKILL.md. This mod also ships 7 executable files (outputs/dsv4_vllm_mi355x-fp4-mtp_agentic_override_dep8_c64/01_docker_run.sh, outputs/dsv4_vllm_mi355x-fp4-mtp_agentic_override_dep8_c64/02_start_server.sh, outputs/dsv4_vllm_mi355x-fp4-mtp_agentic_override_dep8_c64/03_run_client.sh, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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/infx-bench-script-generator/SKILL.md · 71 lines

How it starts

The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.

InferenceX 本地 Benchmark 脚本生成器

根据一个 InferenceX srt-slurm 单机 recipe（xxx.yaml）+ 一个 override variant，生成三份在本地 GPU 服务器上手动跑 AgentX benchmark 的脚本（不依赖 Slurm / srt-slurm / GitHub Actions）。

输入（缺失项向用户确认，路径可留空走默认）

recipe 路径（如 dsv4/vllm/mi355x-fp4-mtp/agentic.yaml，相对于 InferenceX 仓库的 inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/，或绝对路径）
override 名（如 override_tp8_c56、override_dep8_c64；也接受 base）
是否采集 torch profiler trace：off（默认）/ prefill（一个完整 prefill step）/ decode（一个完整 decode step）
模型权重 host 路径（留空 = server 脚本在容器内下载，存 named volume）
InferenceX 仓库 host 路径（留空 = client 脚本在容器内 clone）
trace 数据集 host 路径（留空 = 挂载 host ~/.cache/huggingface，复用缓存与 HF token）
测试结果 host 目录（留空 = named volume 挂到 /logs，即 InferenceX 默认结果位置）

工作流程

读 recipe 并自检：读目标 yaml，确认 override 存在；需要 schema/merge 规则细节时读 references/recipe-schema.md（含 dsv4 mi355x 的 worked example）。

运行生成器（产物由脚本确定性生成，不要手写这三个文件）：

python3 <skill_dir>/scripts/gen_bench_scripts.py \
  --recipe <recipe 绝对路径> --override <override 名> \
  [--model-path <host 路径>] [--repo-path <host 路径>] \
  [--trace-path <host 路径>] [--results-dir <host 路径>] \
  [--profile off|prefill|decode] \
  --out-root <skill_dir>/outputs

PyYAML 缺失时报错，先 python3 -m pip install pyyaml。

复核生成产物（outputs/<recipe路径以_分隔>_<override>/）：
- variant.json / variant.env 与三份脚本一致（server args 与 roles.agg.args、 client env 与 benchmark.env、CONC 与 variant 一致）；
- 01_docker_run.sh 中没有任何 -e 环境变量（env 全部在 02/03 脚本内）；
- 03_run_client.sh 里 IS_AGENTIC=1 等必须先于 source runtime_settings.sh，功耗开关（ENABLE_AGENTX_POWER=0 等）必须在 source 之后。核对领域细节时读 references/server-runbook.md 与 references/client-runbook.md。

汇报：三份脚本的路径、容器名，以及执行顺序：

1. bash outputs/<文件夹>/01_docker_run.sh            # host 上起容器
2. docker exec -it <容器名> bash /infx-scripts/02_start_server.sh
3. docker exec -it <容器名> bash /infx-scripts/03_run_client.sh

Read the full file on GitHub · 71 lines

Files

What ships with it

17 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.

27d ago First seen · 71 lines · 0 tokens per session scan A 0f7b7fa8cfce

Subscribe to this mod's changes

infx-bench-script-generator is a skill published in the GitHub repository shen-shanshan/vllm-dev-skills (22 stars, last pushed yesterday), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,516 tokens. 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 2026-09-29.

Analytics and metrics Claude Code Containers DevOps and cloud agent-skills software-engineering vLLM vibe-coding

moai-ref-secops

DevSecOps, container, and API operational defensive security reference: CI/CD pipeline hardening, secret scanning, IaC misconfiguration detection, SAST/DAST integration, container image scanning, Kubernetes RBAC hardening, container-escape defense, runtime threat detection, OWASP API Top 10 operational defense, WAF…

modu-ai/moai-adk · 201 tokens

memstack-deployment-docker-setup

Use this skill when the user says 'Docker', 'Dockerfile', 'docker-compose', 'containerize', 'docker-setup', or needs to containerize an application with optimized Docker images and compose configurations. Do NOT use for serverless or static site deployments.

cwinvestments/memstack · 62 tokens

containerization-assistant

Generate Dockerfiles, Docker Compose configurations, and Kubernetes manifests for containerizing applications. Use when: (1) Creating Dockerfiles for Node.js, Python, Java, Go, or other applications, (2) Setting up multi-service environments with Docker Compose, (3) Generating Kubernetes deployments, services, and…

ArabelaTso/Skills-4-SE · 105 tokens

vibe-service-health-dashboard

Queries all configured services for health status — uptime, restarts, errors, resources. Use when checking the health of running services.

ash1794/vibe-engineering · 32 tokens

supply-chain-attack-recon

External recon for software supply-chain attack surface — package-namespace squatting candidates, dependency-confusion vulnerabilities, GitHub Actions injection openings, container image registry exposure, SBOM mining, internal-package-name leakage, and CI/CD configuration exposure. Reconnaissance and identification…

adriannoes/awesome-agentic-ai · 141 tokens

performing-container-image-hardening

This skill covers hardening container images by minimizing attack surface, removing unnecessary packages, implementing multi-stage builds, configuring non-root users, and applying CIS Docker Benchmark recommendations to produce secure production-ready images.

adriannoes/awesome-agentic-ai · 46 tokens