Agent Skills › MigoXLab/LMeterX › lmeterx-llm-loadtest

lmeterx-llm-loadtest

GitHub

用于对LLM API（如OpenAI/Claude兼容接口）进行负载测试。自动识别URL或curl命令，执行脚本预检连通性并创建压测任务，支持自定义并发、时长等参数。

.openclaw/skills/lmeterx-llm-loadtest/SKILL.md MigoXLab/LMeterX

Trigger Scenarios

用户提供以 /v1/chat/completions 或 /v1/messages 结尾的 LLM API URL 用户提及 LLM、大模型、OpenAI、Claude 等关键词用户提供包含 model 和 messages 字段的 curl 命令

Install

npx skills add MigoXLab/LMeterX --skill lmeterx-llm-loadtest -g -y

More Options

Non-standard path

npx skills add https://github.com/MigoXLab/LMeterX/tree/main/.openclaw/skills/lmeterx-llm-loadtest -g -y

Use without installing

npx skills use MigoXLab/LMeterX@lmeterx-llm-loadtest

指定 Agent (Claude Code)

npx skills add MigoXLab/LMeterX --skill lmeterx-llm-loadtest -a claude-code -g -y

安装 repo 全部 skill

npx skills add MigoXLab/LMeterX --all -g -y

预览 repo 内 skill

npx skills add MigoXLab/LMeterX --list

SKILL.md

Frontmatter

{
    "name": "lmeterx-llm-loadtest",
    "emoji": "🤖",
    "requires": {
        "env": [
            "LMETERX_BASE_URL"
        ]
    },
    "triggers": [
        "压测这个LLM API",
        "压测这个大模型接口",
        "压测这个模型API",
        "帮我压测 OpenAI",
        "帮我压测 Claude API",
        "压测 chat completions",
        "压测 messages 接口",
        "load test this LLM API",
        "load test OpenAI endpoint",
        "stress test this model API",
        "压测这个AI接口"
    ],
    "description": "LMeterX LLM API Load Test tool. When a user provides an **LLM API endpoint URL**\n(ending with `\/v1\/chat\/completions` or `\/v1\/messages`) or a curl command targeting\nan LLM API, this skill executes a script to pre-check connectivity and create\na load testing task. Supports OpenAI-compatible and Claude-compatible APIs.\n"
}

Skill: lmeterx-llm-loadtest

Intent Routing Rules (Highest Priority)

When to USE this Skill

URL ends with /v1/chat/completions (OpenAI-compatible)
URL ends with /v1/messages (Claude/Anthropic-compatible)
User mentions "LLM", "大模型", "模型接口", "OpenAI", "Claude", "chat completions"
curl command body contains "model" and "messages" fields

When NOT to use this Skill

Condition	Use Instead
URL is a webpage (e.g. `https://www.baidu.com`)	`lmeterx-web-loadtest`
URL is a regular API without LLM path patterns (e.g. `/api/users`, `/graphql`)	`lmeterx-http-loadtest`
User says "网站/网页/页面"	`lmeterx-web-loadtest`

Quick Decision Rule

URL ends with /v1/chat/completions or /v1/messages → THIS SKILL
URL is a normal webpage                            → lmeterx-web-loadtest
Everything else (REST API, GraphQL, etc.)          → lmeterx-http-loadtest

Execution Rules

Mandatory: You must and may only execute the provided script via Bash.
Prohibition: Do NOT manually construct HTTP requests using curl or requests to call LMeterX APIs.
Prohibition: Do NOT fabricate results. Execute the script and respond based on actual stdout output.

The Only Correct Way to Execute

With URL:

export LMETERX_AUTH_TOKEN="${LMETERX_AUTH_TOKEN:-<YOUR_AUTH_TOKEN>}"
python "${SKILL_DIR}/scripts/run.py" \
  --url "<LLM API URL>" \
  --header "Authorization: Bearer <api-key>"

With curl command:

export LMETERX_AUTH_TOKEN="${LMETERX_AUTH_TOKEN:-<YOUR_AUTH_TOKEN>}"
python "${SKILL_DIR}/scripts/run.py" \
  --curl '<full curl command>'

With custom load parameters:

export LMETERX_AUTH_TOKEN="${LMETERX_AUTH_TOKEN:-<YOUR_AUTH_TOKEN>}"
python "${SKILL_DIR}/scripts/run.py" \
  --url "<LLM API URL>" \
  --header "Authorization: Bearer <api-key>" \
  --body '{"model":"gpt-4","messages":[{"role":"user","content":"Hello"}]}' \
  --concurrent-users 50 \
  --duration 300 \
  --spawn-rate 30

Parameters

Parameter	Default	Description
`--url`	(required, or use --curl)	LLM API endpoint URL
`--curl`	(required, or use --url)	Full curl command string
`--header`	[]	Request header (repeatable, format: `Key: Value`)
`--body`	""	Request body JSON string
`--model`	(auto-extracted from body)	Model name
`--stream` / `--no-stream`	true	Enable/disable streaming
`--concurrent-users`	50	Concurrent users (1-5000)
`--duration`	300	Duration in seconds (1-172800)
`--spawn-rate`	30	User spawn rate
`--name`	(auto-generated)	Task name
`--test-data`	""	Dataset: `""` for none, `"default"` for built-in dataset

Presenting Results to the User

After execution, present:

API Type: OpenAI Chat or Claude Chat
Pre-check Result: Pass/Fail with categorized failure reason
Task ID and Report URL: {LMETERX_BASE_URL}/results/{task_id}

Exception Handling

Error Scenario	Output Message
HTTP 401/403	LMeterX token is invalid or expired; check `LMETERX_AUTH_TOKEN`
HTTP 5xx	LMeterX platform service error; try again later
Connection Failure	Cannot connect to LMeterX service; check network
Target API 401	Target LLM API requires auth; check Authorization header
Target API timeout	Target LLM API timed out; may be overloaded