← Back to Skills Marketplace
tatekii

AIRILAB

by Siyn Ma · GitHub ↗ · v0.0.1 · MIT-0
cross-platform ✓ Security Clean
96
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install airilab
Description
统一调用 AiriLab 图像生成与任务管理能力(MJ 创意渲染 workflowId=0、创意放大 workflowId=16、氛围转换 workflowId=13),并自动处理登录鉴权、项目选择、任务提交、状态轮询与结果回推。用户提到 Midjourney/MJ 文生图、参考图生成、放大超分/4K、天气季节昼...
README (SKILL.md)

AiriLab Skill

目标

  • 将 AiriLab SaaS 的多接口流程封装为可对话调用的单一技能。
  • 在调用图像能力前,自动保证登录态与项目上下文有效。
  • 通过后台 worker 追踪异步任务并主动回推结果。

能力范围

  • 鉴权:手机号 + 验证码登录,持久化 token。
  • 项目管理:拉取团队/项目、选择项目、持久化项目上下文。
  • 图像工作流:
    • workflowId=0 MJ 创意渲染(文生图/参考图生成)
    • workflowId=16 创意放大
    • workflowId=13 氛围转换(天气/季节/光照语义转换)
  • 异步任务:提交后轮询状态,完成后拉取记录并通知用户。

触发规则

当用户表达以下任一意图时自动触发:

  • MJ 创意渲染:文生图、概念图、效果图、参考图生成、多图出图。
  • 创意放大:清晰度增强、超分辨率、长边 4K 放大、细节/纹理增强。
  • 氛围转换:天气/季节/昼夜切换,提升画面氛围、故事感和视觉张力。
  • AiriLab 流程操作:登录、验证码、选择项目、切换项目、查询任务状态、拉取任务结果。

强触发关键词(任一命中即可):

  • MJ/渲染类:mjmidjourney文生图渲染概念图效果图参考图出图生成4张
  • 放大类:创意放大放大高清超分4k长边4k细节增强纹理增强
  • 氛围类:氛围转换atmosphere天气切换季节切换白天转夜景晴天雨天雪景
  • 流程类:airilab登录验证码登录选择项目切换项目projectIdjobId任务状态查询结果

高置信触发语句示例:

  • “用 MJ 生成 4 张现代建筑外立面概念图。”
  • “把这张图做创意放大,长边 4K,细节和质感增强。”
  • “把这张白天街景转成雨夜氛围,保留构图。”
  • “帮我登录 AiriLab,然后列出项目让我选一个。”
  • “这个 jobId 帮我查下状态,完成后把结果发我。”

触发优先级:

  1. 用户上传图像且包含“放大/氛围转换/改图”语义 -> 优先匹配 workflowId=16workflowId=13
  2. 无上传图像但包含“生成/渲染/MJ”语义 -> 匹配 workflowId=0
  3. 包含“登录/项目/job 状态/结果”语义 -> 进入鉴权、项目或异步任务查询流程。

不触发条件(避免误触发):

  • 用户仅讨论通用设计理论、审美评价,不要求生成/改图/放大/状态查询。
  • 用户请求与 AiriLab 无关的纯文本任务(翻译、总结、写作)且未提及图像工作流。

前置条件(强约束)

  1. 调用任何 AiriLab 接口前必须检查登录态。
  2. 提交图像任务前必须存在已选项目(teamId/projectId/projectName)。
  3. 登录失效时必须引导手机号验证码登录,不要求用户提供 token。

工作流参数约束

1) MJ 创意渲染(workflowId=0

  • 必填:prompt
  • 可选:referenceImage(最多 3 张)、imageCount(1-4)

2) 创意放大(workflowId=16

  • 必填:baseImage
  • 可选:widthheight

3) 氛围转换(workflowId=13

  • 必填:baseImage
  • 可选:promptreferenceImage(最多 1 张)、imageCount

多语言文案规范

以下文案用于能力介绍展示,字段名固定,换行统一使用 \

1) MJ 创意渲染(workflowId=0

desc: "The most creative and aesthetic AI model for fast, high-quality concept visuals."
descCN: "美学和创意性最强的AI模型,快速生成高质量概念意向图"
descFR: "Le modèle d’IA le plus créatif et esthétique pour générer rapidement des visuels conceptuels de haute qualité."

2) 创意放大(workflowId=16

content: "4K Long-Edge Upscale with Enhanced Details and Texture"
contentCN: "放大至长边4K\
增强细节和质感"
contentFR: "Agrandissement 4K (côté long) avec amélioration des détails et des textures"

3) 氛围转换(workflowId=13

heading: "Atmosphere swift"
headingCN: "氛围转换"
headingFR: "Atmosphère rapide"
content: "Quickly switch between weather and seasons\
Applicable scenarios: Rapidly enhance atmospheric expression, improve the narrative and visual tension of the proposal."
contentCN: "快速切换画面天气、季节\
适用场景:快速提升氛围表达,提升方案故事感与视觉张力"
contentFR: "Changez rapidement entre la météo et les saisons\
Scénarios applicables : Améliorez rapidement l'expression atmosphérique, améliorez la narration et la tension visuelle de la proposition."

异步任务处理(强约束)

轮询任务状态与查询生成记录必须分离:

  • 轮询状态接口:GET /api/Universal/Job/{jobId}
    • 状态字段来源:data.status
  • 记录查询接口:POST /api/CrudRouters/getOneRecord
    • 仅在状态为 completed 后调用

脚本职责:

  • scripts/check_status.py:只做状态轮询,输出 status:\x3Cvalue>
  • scripts/fetch.py:只做记录查询,返回生成 URL 列表等结果。
  • scheduler/worker.py:负责任务生命周期与通知。
  • 任务提交成功后必须立即结束本轮对话,并明确告知用户“后台完成后会自动通知”。

本地持久化

  • 运行根目录:优先使用环境变量 AIRILAB_HOME,未设置时默认 ~/.openclaw/skills/airilab
  • config/.env:token 与登录相关信息。
  • config/project_config.json:当前项目上下文。
  • scheduler/jobs.db:异步任务状态。
  • scheduler/worker.log:worker 运行日志。
  • scheduler/worker.pid:worker 单实例锁文件。

运行命令

# 安装依赖
python -m pip install -r ~/.openclaw/skills/airilab/requirements.txt

# 查看配置状态
python ~/.openclaw/skills/airilab/core/config.py status

# 查看运行健康状态(token/project/worker/jobs)
python ~/.openclaw/skills/airilab/core/config.py health

# 一键健康检查(状态 + worker + 最近日志)
~/.openclaw/skills/airilab/scripts/health.sh

# 安装后初始化(依赖 + worker + 自启动)
~/.openclaw/skills/airilab/scripts/post-install.sh

# 启动后台 worker
~/.openclaw/skills/airilab/scripts/start-worker.sh

故障处理

  • missing_token:引导用户登录。
  • missing_project:引导用户选择项目。
  • timeout / 轮询超时:任务标记失败并通知用户重试。
  • 记录查询 success=false:按失败处理,不得误判成功。

相关文件

  • core/config.py
  • core/auth.py
  • core/project.py
  • core/upload.py
  • core/api.py
  • scripts/check_status.py
  • scripts/fetch.py
  • scheduler/worker.py
  • SPEC.md\r

Agent 执行约束(强约束)

1) 执行责任

  • Agent 必须直接执行可执行步骤,不得把可执行命令转嫁给用户。
  • 禁止输出“你可以手动执行…/请你在终端执行…”这类兜底话术,除非明确出现权限/环境阻塞且无法由 Agent 代执行。
  • 登录场景下,Agent 只向用户索取最小必要输入:phonecode;拿到后立即自行执行登录命令。

2) 路径解析与工作目录

  • skill_root:以当前 SKILL.md 所在目录为准,不依赖对话中的硬编码路径。
  • runtime_root:优先 AIRILAB_HOME,未设置时使用 ~/.openclaw/skills/airilab
  • Agent 执行命令时必须使用绝对路径,且 cwd 设置为 skill_root,避免 cd 到错误目录。

3) 命令调用规范

  • 登录:python core/auth.py login --phone \x3Cphone> --code \x3Ccode>(由 Agent 执行)
  • 状态:python core/config.py status(由 Agent 执行)
  • 健康:python core/config.py healthscripts/health.sh(由 Agent 执行)
  • 启动 worker:scripts/start-worker.sh(由 Agent 执行)

4) 对话输出规范

  • 当缺少参数时,只提问缺失参数,不提供“手动执行命令”替代方案。
  • 执行后返回结果摘要(成功/失败、关键字段、下一步),不要只返回命令文本。
  • 若必须让用户介入,只允许两类:输入验证码、确认业务选择(例如项目选择)。\r
Usage Guidance
This skill appears to do what it claims (manage AiriLab jobs) and stores only AiriLab tokens and project configs locally. Before installing or running any post-install scripts: 1) review and consent to storing your AiriLab token under ~/.openclaw/skills/airilab/config/.env; 2) be aware post-install can install Python packages and enable a user-level service or crontab to run a background worker — if you don't want a persistent background process, do not run post-install.sh or pass --no-autostart and avoid enabling the systemd/crontab steps; 3) the agent will ask you for phone + OTP to login — supply those only if you trust the skill and account flow; 4) if you want higher assurance, inspect/execute the scripts manually rather than allowing automatic agent execution. If you want me to highlight exact lines that create autostart or persist tokens, say so and I will point them out.
Capability Analysis
Type: OpenClaw Skill Name: airilab Version: 0.0.1 The AiriLab skill bundle is a legitimate integration for AI image generation and task management. It implements a standard authentication flow (OTP via phone), project selection, and an asynchronous task polling system. While the bundle includes persistence mechanisms via systemd and crontab (found in `scripts/post-install.sh` and `scripts/setup-autostart.sh`) to maintain a background worker process, these actions are explicitly documented in `SKILL.md` and are necessary for the stated purpose of tracking long-running image generation jobs. The code follows secure practices for internal script execution and limits data access to its own configuration directory (`~/.openclaw/skills/airilab`).
Capability Assessment
Purpose & Capability
Name/description match the code and SKILL.md: the package implements AiriLab image generation (MJ/upscale/atmosphere), authentication (phone+OTP), project selection, task submission, polling and result fetching. Required resources (local filesystem, network to cn.airilab.com) are coherent with that purpose.
Instruction Scope
SKILL.md gives precise runtime instructions: the agent should run local python scripts to login, check status, fetch results and start a background worker. It will ask the user only for phone and OTP for login (per the constraints). This is within scope for a wrapper that manages async SaaS jobs, but it requires the agent to execute commands and receive OTPs from the user — a privacy/usability consideration the user should accept explicitly.
Install Mechanism
Registry lists no formal install spec, but the bundle includes post-install.sh which runs pip install -r requirements.txt and configures autostart (systemd user service or cron). The dependency install is simple (requests only) and all downloads go to the Python package index via pip; no unknown remote archives or URL shorteners are used. Still, running post-install.sh will modify user-level services/crontab and install packages locally.
Credentials
The skill declares no required env vars, and the code only optionally uses AIRILAB_HOME (defaults to ~/.openclaw/skills/airilab). It stores the AiriLab token locally in config/.env and project info in project_config.json. It does not request unrelated cloud credentials or external secrets. The phone/OTP interaction is necessary for the stated authentication flow.
Persistence & Privilege
The package includes scripts (post-install.sh, setup-autostart.sh, start-worker.sh) that can enable a user-level systemd service or add a crontab @reboot entry and will start a long-running worker process that polls AiriLab and writes files under the user's home (~/.openclaw). While this persistence is coherent with background job polling, it does grant the skill continuous background presence on the host if the user runs the post-install/autostart steps. Consider whether you want that runtime persistence before enabling autostart.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install airilab
  3. After installation, invoke the skill by name or use /airilab
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.0.1
AiriLab Skill v1.0.0 - Unified entry for AiriLab’s image workflows: creative MJ rendering, 4K upscale, and atmosphere/season conversion, all with integrated login, project management, and task state tracking. - Automatic login and project context handling before any operation; only minimal user input (phone/code) is ever requested. - Asynchronous tasks are submitted, status is polled in the background, and user is notified automatically when complete. - Strong trigger rules based on user intent and keywords for image generation, upscaling, atmosphere conversion, and workflow operations (login, project, job status). - Robust error handling, separation of polling and record-fetching functions, and clear user feedback on missing tokens, projects, or task failures. - Command-line tools and scripts included for setup, health checks, and worker management, with clear execution responsibilities for agents.
Metadata
Slug airilab
Version 0.0.1
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is AIRILAB?

统一调用 AiriLab 图像生成与任务管理能力(MJ 创意渲染 workflowId=0、创意放大 workflowId=16、氛围转换 workflowId=13),并自动处理登录鉴权、项目选择、任务提交、状态轮询与结果回推。用户提到 Midjourney/MJ 文生图、参考图生成、放大超分/4K、天气季节昼... It is an AI Agent Skill for Claude Code / OpenClaw, with 96 downloads so far.

How do I install AIRILAB?

Run "/install airilab" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is AIRILAB free?

Yes, AIRILAB is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does AIRILAB support?

AIRILAB is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created AIRILAB?

It is built and maintained by Siyn Ma (@tatekii); the current version is v0.0.1.

💬 Comments