Ai Image Cli

AI 图像生成工具。支持文生图(text2img)和图生图(img2img)功能，基于 AIGW API。 激活场景：生成图片、画图、制作图像、修改图片、图片编辑、 或提到 "生成图片"、"画一个"、"创建图像"、"修改这张图"、"改成"、"转为图片" 时激活。

Security Scan

The skill claims to be a simple AI image CLI, but its instructions and helper scripts expect/consume credentials and configuration files that are not declared and install from an internal PyPI (with embedded credentials), so the package footprint does not fully match the published metadata.

Purpose & Capability

The SKILL.md describes an image-generation CLI that requires an AIGW API key (LANGBASE_TOKEN or ARK_API_KEY), and the shipped docs/scripts clearly depend on those tokens — yet the skill registry metadata declares no required environment variables or config paths. That mismatch (claimed no secrets required vs. instructions expecting service tokens and a user.env file) is a substantive incoherence.

Instruction Scope

Runtime instructions and helper scripts instruct the agent to source ~/.openclaw/user.env and /home/appops/.openclaw/user.env to obtain LANGBASE_TOKEN/ARK_API_KEY, run preflight/install scripts, and call the CLI. Those instructions cause the agent to read local config files and environment variables not described in the manifest; they also recommend running pip to install a package from a private index. The SKILL.md explicitly tells the user/agent to load and rely on local user.env files — this expands scope beyond a pure, self-contained instruction-only skill.

Install Mechanism

No install spec was registered, but scripts/install.sh will pip-install the package, defaulting to a private PyPI host (music-pypi.hz.netease.com). The script contains a DEFAULT_PYPI_INDEX value embedding basic-auth credentials (http://avlab:[e-mail]/simple). Pulling code from an internal/non-standard index (and with hard-coded credentials) is higher risk and unexpected for a publicly listed skill.

Credentials

The code and documentation require LANGBASE_TOKEN or ARK_API_KEY and instruct sourcing user.env files to obtain them, but the registry declares no required env vars or config paths. The install script also reads and uses PIP_INDEX_URL / PIP_TRUSTED_HOST (and falls back to an index with embedded credentials). Requesting access to local user.env files and expecting service tokens without declaring them is disproportionate and opaque.

Persistence & Privilege

The skill does not request always: true and does not claim elevated platform privileges. However, its install script will persist system/user-level Python packages (pip install --user or global), which alters the environment and leaves installed code on disk. This is not automatically malicious but is a notable persistent effect the user should accept explicitly.

What to consider before installing

This skill looks like a wrapper for a private AIGW-backed image CLI but is inconsistent with its registry metadata. Before installing or running it: (1) don't run the install.sh or preflight.sh blindly — they will source ~/.openclaw/user.env and may install code from a private PyPI (the script even includes baked-in credentials); (2) ask the publisher for a declared list of required env vars and config paths (LANGBASE_TOKEN / ARK_API_KEY and the exact user.env format) and for an explanation of the private PyPI usage; (3) inspect the ai-image-cli package source you would install (or prefer to install from a trusted public repository) and verify the package maintainer; (4) if you must test, run preflight/install in an isolated environment (container or VM) with no sensitive credentials mounted; (5) if any credentials (e.g., avlab:avlab123) in the scripts are real for your environment, consider them compromised and rotate them. If the publisher can provide a manifest that declares the required env vars and justifies the private-index install, the concerns here would be reduced.

AI Image — AI 图像生成技能

ai-image-cli 是 AI 图像生成工具，支持文生图和图生图两大核心能力。

能力	提供商	模型	说明
文生图	AIGW	doubao-seedream-4-5	根据文本描述生成图片
图生图	AIGW	doubao-seededit-3-0-i2i	基于源图片和文本生成新图片

命令选择

根据用户意图，直接选择对应子命令：

• 用户想根据描述生成新图片 → ai-image text2img "<描述>"
• 用户提供了图片 URL，想要修改或编辑 → ai-image img2img "<修改描述>" <URL>
• 查看工具能力 → ai-image capabilities

命令速查

命令	用途	关键选项
ai-image text2img "<描述>"	文生图	-s 尺寸, --seed 种子, -g 引导系数, --watermark
ai-image img2img "<描述>" <URL>	图生图	--seed 种子, -g 引导系数, --watermark
ai-image capabilities	查看工具能力(JSON)	无参数，无需认证

所有命令输出 JSON 格式，success 字段标识成功或失败。

Quick Start

直接执行命令即可，无需预检。如果命令不存在或返回错误，再执行环境诊断：

# 诊断环境
bash {baseDir}/scripts/preflight.sh

# 如果 cli_installed: false，执行安装
bash {baseDir}/scripts/install.sh

常用工作流

文生图 (text2img)

# 基本用法
ai-image text2img "一个中国女孩,高清"

# 指定尺寸
ai-image text2img "夕阳下的海滩" --size 1920x1920

# 使用随机种子(可复现)
ai-image text2img "雪山风景" --seed 12345

# 完整参数
ai-image text2img "城市夜景" --size 1920x1920 --seed 999 --guidance-scale 3.0 --watermark

常用参数：

• --size, -s: 图片尺寸，如 1920x1920(默认: 1920x1920，最低尺寸：1920x1920)
• --seed: 随机种子，用于复现相同结果
• --guidance-scale, -g: 引导系数，控制生成质量(默认: 2.5)
• --watermark / --no-watermark: 是否添加水印(默认: 不添加)

图生图 (img2img)

注意: 图生图不支持指定尺寸参数，输出图片尺寸由源图片决定。

# 基本用法
ai-image img2img "改成爱心形状的泡泡" https://example.com/image.jpg

# 使用随机种子
ai-image img2img "改变颜色为红色" https://example.com/car.jpg --seed 777

# 调整引导系数
ai-image img2img "转为水彩画风格" https://example.com/pic.png --guidance-scale 6.0

# 完整参数
ai-image img2img "添加蓝天白云" https://example.com/photo.png \
  --seed 888 --guidance-scale 6.0 --watermark

常用参数：

• --seed: 随机种子，用于复现相同结果
• --guidance-scale, -g: 引导系数，控制编辑强度(默认: 5.5)
• --watermark / --no-watermark: 是否添加水印(默认: 不添加)

输出格式

成功响应

{
  "success": true,
  "command": "text2img",
  "provider": "ark",
  "data": {
    "model": "doubao-seedream-4-5",
    "created": 1754384045,
    "data": [
      {
        "url": "https://ark-content-generation-v2-cn-beijing.tos-cn-beijing.volces.com/..."
      }
    ],
    "usage": {
      "generated_images": 1,
      "output_tokens": 4096,
      "total_tokens": 4096
    }
  }
}

错误响应

{
  "success": false,
  "command": "text2img",
  "error": "未找到 AIGW API Key",
  "hint": "请设置 LANGBASE_TOKEN 或 ARK_API_KEY 环境变量"
}

故障恢复

命令执行失败时，按以下步骤排查：

1. 命令不存在 (command not found): 执行 bash {baseDir}/scripts/install.sh
2. 认证失败 (AUTH_FAILED 或 未找到 AIGW API Key): 执行 bash {baseDir}/scripts/preflight.sh 查看诊断。若 auth_configured: false，说明 LANGBASE_TOKEN 环境变量未注入当前进程，需确认 ~/.openclaw/user.env 中包含 LANGBASE_TOKEN。
3. 网络/API 错误: 重试一次。若持续失败，执行 preflight 检查连通性。

未全局安装时的备选方式

python3 -m ai_image text2img "查询词"
python3 -m ai_image img2img "修改描述" "https://example.com/image.jpg"

参考文档

• 完整 CLI 参考（参数、返回结构、错误码）：读取 references/cli-reference.md
• 更多高级示例（批量生成、参数调优）：读取 references/examples.md