Agent skill

Bench

构建并运行 vlink-bench 性能基准(showcase/quick/full 预设),生成 HTML/JSON 报告。用户要求"跑 bench"、"性能测试"、对比后端吞吐/延迟、验证性能回归时使用。

Apache-2.0Auto-check passed

Install Bench

skills CLI

$ npx skills add thun-res/vlink --skill bench -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI

$ gh skill install thun-res/vlink bench --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy

$ git clone --depth 1 https://github.com/thun-res/vlink.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/bench .claude/skills/bench && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Paste into Claude Code

Install the "bench" agent skill from https://github.com/thun-res/vlink/tree/master/.agents/skills/bench into .claude/skills/bench/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "bench", then confirm the skill loads.

Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

$skill-installer

$skill-installer install https://github.com/thun-res/vlink/tree/master/.agents/skills/bench

Type this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.

skills CLI

$ npx skills add thun-res/vlink --skill bench -a codex

Project install goes to .agents/skills/; add -g for ~/.codex/skills/.

GitHub CLI

$ gh skill install thun-res/vlink bench --agent codex

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/thun-res/vlink.git skills-src && mkdir -p .agents/skills && cp -r skills-src/.agents/skills/bench .agents/skills/bench && rm -rf skills-src

Use ~/.agents/skills/ instead of .agents/skills for a personal install.

Codex skills documentation · loads skills from .agents/skills/

Paste into Codex

Install the "bench" agent skill from https://github.com/thun-res/vlink/tree/master/.agents/skills/bench into .agents/skills/bench/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "bench", then confirm the skill loads.

Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

skills CLI

$ npx skills add thun-res/vlink --skill bench -a cursor

Project install goes to .agents/skills/; add -g for ~/.cursor/skills/.

GitHub CLI

$ gh skill install thun-res/vlink bench --agent cursor

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/thun-res/vlink.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/.agents/skills/bench .cursor/skills/bench && rm -rf skills-src

Use ~/.cursor/skills/ instead of .cursor/skills for a personal install.

Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/

Paste into Cursor

Install the "bench" agent skill from https://github.com/thun-res/vlink/tree/master/.agents/skills/bench into .cursor/skills/bench/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "bench", then confirm the skill loads.

Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

gemini skills install

$ gemini skills install https://github.com/thun-res/vlink.git --path .agents/skills/bench

--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.

skills CLI

$ npx skills add thun-res/vlink --skill bench -a gemini-cli

Project install goes to .agents/skills/; add -g for ~/.gemini/skills/.

GitHub CLI

$ gh skill install thun-res/vlink bench --agent gemini-cli

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/thun-res/vlink.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/.agents/skills/bench .gemini/skills/bench && rm -rf skills-src

Use ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.

Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/

Paste into Gemini CLI

Install the "bench" agent skill from https://github.com/thun-res/vlink/tree/master/.agents/skills/bench into .gemini/skills/bench/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "bench", then confirm the skill loads.

Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

GitHub CLI

$ gh skill install thun-res/vlink bench

Installs for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).

skills CLI

$ npx skills add thun-res/vlink --skill bench -a github-copilot

Project install goes to .agents/skills/; add -g for ~/.copilot/skills/.

Manual copy

$ git clone --depth 1 https://github.com/thun-res/vlink.git skills-src && mkdir -p .github/skills && cp -r skills-src/.agents/skills/bench .github/skills/bench && rm -rf skills-src

Use ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.

GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/

Paste into GitHub Copilot

Install the "bench" agent skill from https://github.com/thun-res/vlink/tree/master/.agents/skills/bench into .github/skills/bench/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "bench", then confirm the skill loads.

GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

skills CLI

$ npx skills add thun-res/vlink --skill bench -a opencode

OpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.

GitHub CLI

$ gh skill install thun-res/vlink bench --agent opencode

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/thun-res/vlink.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/.agents/skills/bench .opencode/skills/bench && rm -rf skills-src

Use ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.

OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/

Paste into OpenCode

Install the "bench" agent skill from https://github.com/thun-res/vlink/tree/master/.agents/skills/bench into .opencode/skills/bench/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "bench", then confirm the skill loads.

OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

Facts

Skill name: bench
GitHub stars: 115
Token cost: ~993 tokens
SKILL.md length: 97 words
Files: 2
Skills in repo: 16
Repo updated: Oct 7, 2026
First seen: Oct 7, 2026
Licence: Apache-2.0

At a glance

构建并运行 vlink-bench 性能基准(showcase/quick/full 预设),生成 HTML/JSON 报告。用户要求"跑 bench"、"性能测试"、对比后端吞吐/延迟、验证性能回归时使用。

Works in 3 steps: 构建 → 运行 → 结果处理
SKILL.md covers 1. 构建, 2. 运行 and 3. 结果处理
Calls cmake and git

What it does

Bench is an agent skill from thun-res/vlink. 构建并运行 vlink-bench 性能基准(showcase/quick/full 预设),生成 HTML/JSON 报告。用户要求"跑 bench"、"性能测试"、对比后端吞吐/延迟、验证性能回归时使用。

Its SKILL.md is about 990 tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

The repository describes itself as: VLink is a high-performance C++ communication middleware for autonomous driving and embodied intelligence, positioned as a full-scenario alternative to ROS 2. The licence is Apache-2.0.

Example prompts

“跑 bench”
“/bench”

Workflow steps

3 steps, taken from the step headings in SKILL.md.

构建
运行
结果处理

What it can do on your machine

Read from SKILL.md and the folder at commit 1793889. It shows what the files ask for, not the result of running them.

Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
- cmake
- git
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Bench loads about 993 tokens when it runs. Until then it costs about 28 tokens; SKILL.md has 97 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it: ~28
When it runs · the whole SKILL.md, loaded when a task matches: ~993

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from thun-res/vlink at commit 1793889, republished under its Apache-2.0 licence (© thun-res). 97 words, ~993 tokens.

Download SKILL.mdSave it as .claude/skills/bench/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.

name: bench
description: 构建并运行 vlink-bench 性能基准(showcase/quick/full 预设),生成 HTML/JSON 报告。用户要求"跑 bench"、"性能测试"、对比后端吞吐/延迟、验证性能回归时使用。

性能基准测试

vlink-bench 是仓库自带的基准 CLI(cli/bench/),由默认开启的 ENABLE_CLI_BENCH=ON 构建,二进制位于 build-ai/skill_bench/output/bin/vlink-bench。官方 Wiki 的基准页即来自 quick 预设的一次运行 (.github/scripts/release-bench.sh)。

1. 构建

Linux / macOS:

bash

REPO_ROOT="$(git rev-parse --show-toplevel)"
BUILD_DIR="$REPO_ROOT/build-ai/skill_bench"
export PYTHONPYCACHEPREFIX="$BUILD_DIR/__pycache__"
PHYSICAL_CORES=
case "$(uname -s 2>/dev/null)" in
  Linux)
    PHYSICAL_CORES="$(
      LC_ALL=C lscpu -p=CORE,SOCKET 2>/dev/null |
        awk -F, '
          $1 !~ /^#/ && $1 ~ /^[0-9]+$/ && $2 ~ /^[0-9]+$/ {
            cores[$2 SUBSEP $1] = 1
          }
          END {
            for (core in cores) {
              count++
            }
            if (count > 0) {
              print count
            }
          }'
    )" || PHYSICAL_CORES=
    ;;
  Darwin)
    PHYSICAL_CORES="$(sysctl -n hw.physicalcpu 2>/dev/null)" || PHYSICAL_CORES=
    ;;
esac
case "$PHYSICAL_CORES" in
  '' | *[!0-9]* | 0) PHYSICAL_CORES=1 ;;
esac
if [ "$PHYSICAL_CORES" -gt 1 ]; then
  BUILD_JOBS=$((PHYSICAL_CORES - 1))
else
  BUILD_JOBS=1
fi
cmake -S "$REPO_ROOT" -B "$BUILD_DIR" \
  -DENABLE_CXX_STD_20=OFF \
  -DENABLE_CLI_BENCH=ON
cmake --build "$BUILD_DIR" --target vlink-bench --parallel "$BUILD_JOBS"

Windows PowerShell:

powershell

$RepoRoot = git rev-parse --show-toplevel
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
$BuildDir = Join-Path (Join-Path $RepoRoot "build-ai") "skill_bench"
$env:PYTHONPYCACHEPREFIX = Join-Path $BuildDir "__pycache__"
try {
    $Processors = @(Get-CimInstance -ClassName Win32_Processor -ErrorAction Stop)
    if ($Processors.Count -eq 0) {
        throw "No processor information"
    }
    $PhysicalCores = 0
    foreach ($Processor in $Processors) {
        $Cores = [int]$Processor.NumberOfCores
        if ($Cores -lt 1) {
            throw "Invalid physical core count"
        }
        $PhysicalCores += $Cores
    }
} catch {
    $PhysicalCores = 1
}
$BuildJobs = [Math]::Max($PhysicalCores - 1, 1)
& cmake -S $RepoRoot -B $BuildDir `
  -DENABLE_CXX_STD_20=OFF `
  -DENABLE_CLI_BENCH=ON
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
& cmake --build $BuildDir --target vlink-bench --parallel $BuildJobs
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

若 build-ai/skill_bench 已配置过、配置仍适用且未被其他任务使用, 可直接执行第二条;否则使用 skill_bench_<task_name>。配置失败必须原样报告,不得继续使用陈旧构建目录或清理其他构建目录。

2. 运行

与 CI 发布报告一致的跑法:

Linux / macOS:

bash

BENCH_REPORT_DIR="$(mktemp -d)"
"$BUILD_DIR/output/bin/vlink-bench" run \
  --preset quick \
  --report html,json \
  --silent \
  -o "$BENCH_REPORT_DIR/vlink-bench-report"

Windows PowerShell:

powershell

$BenchReportDir = Join-Path ([System.IO.Path]::GetTempPath()) (
  "vlink-bench-" + [guid]::NewGuid()
)
New-Item -ItemType Directory -Path $BenchReportDir | Out-Null
$Bench = Join-Path $BuildDir "output\bin\vlink-bench.exe"
$Output = Join-Path $BenchReportDir "vlink-bench-report"
& $Bench run --preset quick --report html,json --silent -o $Output

--preset:showcase(默认,演示)/ quick(CI 用,较快)/ full(完整矩阵,耗时长)。
--mode:运行形态 local-direct、local-loop 或 process; 传输后端通过 --url 选择。
-o/--output:报告文件前缀(不含扩展名),生成 .html / .json。
每次使用独立 mktemp 目录,避免并发运行或重复执行覆盖报告;完成后报告该目录,由用户决定保留或删除。
其他子命令:plot(由 JSON 重绘报告)、pub/sub(跨进程手动压测端)。完整参数见 vlink-bench run --help。

3. 结果处理

基准数据受宿主机负载影响大:对比性能回归时,before/after 必须在同一台空闲机器、同一预设下运行;正式对比使用 full --repeat 3。
$BUILD_JOBS / BuildJobs 必须严格等于 max(真实物理核心数 - 1, 1)。禁止改用逻辑 CPU 数、裸 --parallel/-j 或固定高并行度;无法可靠获取时固定单核,且同一时刻只运行一个本地构建,防止编译卡死或耗尽内存。
涉及性能结论时,引用 JSON 报告中的具体指标(吞吐/延迟分位数),不凭单次 HTML 观感下结论。
退出码 2 表示运行和报告生成完成,但存在失败 case;仍需读取 JSON 定位失败项,不得误报为命令执行故障。其他非零退出码按执行失败处理。
只在用户显式调用本 skill 时才执行构建与运行;日常改码流程仍遵循 AGENTS.md 强制规则第 3 条(不主动构建)。

Files

SKILL.md and 1 other file in .agents/skills/bench of thun-res/vlink.

SKILL.md
agents/openai.yaml

Open the folder on GitHubat commit 1793889

Compare with similar skills

Bench next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Bench compared with similar skills
Skill	Stars	Used in	Tokens	Auto-check	Licence	Repo updated
Bench this skillthun-res/vlink	115	—	~993	Automated safety check: Pass	Apache-2.0	today
Harness Benchruvnet/ruflo	74k	1 repos	~586	Automated safety check: Notes	MIT	today
Sandbox Benchvercel/next.js	143k	—	~4.1k	Automated safety check: Pass	MIT	today
Bench Readgithub/awesome-copilot	40k	—	~747	Automated safety check: Pass	MIT	today
Benchddalcu/mlx-serve	1.8k	1 repos	~1.1k	Automated safety check: Pass	Custom licence	today
Terminal Bench Looppaperclipai/paperclip	99k	—	~6.3k	Automated safety check: Pass	MIT	today

Similar skills

Harness Bench
ruvnet/ruflo
Manage @metaharness/darwin bench suites — bench create <repo scaffolds a JSON suite from a repo's test corpus; bench verify <suite.json checks suite well-formedness.
74k GitHub starsUsed in 1 repo~586 tokens
DevelopmentAuto-check: notes
Sandbox Bench
vercel/next.js
Official
Benchmark React or Next.js changes on Vercel Sandbox VMs with paired A/B statistics: react PR/commit vs base, or Next.js PR/commit vs base, measured end-to-end through the bench/render-pipeline app…
143k GitHub stars~4.1k tokensUpdated today
Data & AnalyticsAuto-check passed
Bench Read
github/awesome-copilot
Official
Read artifacts from the shared bench — the workspace where desks leave findings, verdicts, and work products for each other and the operator.
40k GitHub stars~747 tokensUpdated today
Auto-check passed
Bench
ddalcu/mlx-serve
mlx-serve benchmarking methodology — bench.sh/llmprobe usage, comparison-trap rules (same-methodology cells only, spec-decode variance, thermal lies, engine naming), perf-claim etiquette.
1.8k GitHub starsUsed in 1 repo~1.1k tokens
Media & CreativeAuto-check passed
Terminal Bench Loop
paperclipai/paperclip
Run one Terminal-Bench task through a bounded Paperclip smoke/diagnosis/fix loop.
99k GitHub stars~6.3k tokensUpdated today
Auto-check passed
Harness Security Bench
ruvnet/ruflo
Run @metaharness/darwin security bench (upstream "Darwin Shield" / ADR-155) — evolves a champion security-detection harness against a 10-vuln / 9-decoy corpus and grades it on…
74k GitHub stars~1.1k tokensUpdated today
DevelopmentAuto-check: notes

Questions about Bench

What does Bench do?

构建并运行 vlink-bench 性能基准(showcase/quick/full 预设),生成 HTML/JSON 报告。用户要求"跑 bench"、"性能测试"、对比后端吞吐/延迟、验证性能回归时使用。. Bench is an agent skill from thun-res/vlink.

How do I install Bench in Claude Code?

Run `npx skills add thun-res/vlink --skill bench -a claude-code`. Or copy the skill folder (.agents/skills/bench in thun-res/vlink) into .claude/skills/bench in your project. Claude Code loads it when a task matches its description.

How do I install Bench in Codex?

Run `npx skills add thun-res/vlink --skill bench -a codex`. Or copy the skill folder (.agents/skills/bench in thun-res/vlink) into .agents/skills/bench in your project. Codex loads it when a task matches its description.

Can I use Bench in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add thun-res/vlink --skill bench -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/bench, .gemini/skills/bench, .github/skills/bench and .opencode/skills/bench in your project.

What does Bench need to run?

Going by SKILL.md and its folder, Bench needs the command-line tools its instructions call (cmake and git).

Does Bench access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Bench safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Bench use?

Bench is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Bench use?

About 993 tokens (SKILL.md is roughly 4k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Bench?

Skills that share tags, products or a category with Bench: Harness Bench (ruvnet/ruflo, 74k stars), Sandbox Bench (vercel/next.js, 143k stars), Bench Read (github/awesome-copilot, 40k stars) and Bench (ddalcu/mlx-serve, 1.8k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Bench?

thun-res (a GitHub user) maintains it in thun-res/vlink, which has 115 GitHub stars. The repository holds 16 skills in this directory. The repository was last updated on October 7, 2026.

Source: thun-res/vlink on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.

Bench

Install Bench

Facts

At a glance

What it does

Example prompts

Workflow steps

What it can do on your machine

Tool permissions

Runs code

Network

Credentials

Context cost

Safety

SKILL.md

性能基准测试

1. 构建

2. 运行

3. 结果处理

Files

Compare with similar skills

Similar skills

Harness Bench

Sandbox Bench

Bench Read

Bench

Terminal Bench Loop

Harness Security Bench

More from thun-res/vlink

Issue

Asan

Cicd

Clang Tidy

Commit

Coverage

Questions about Bench