Sih Ai Photo Editor

AI图片编辑工具，支持自然语言驱动的换装、换背景、换脸、风格转换（动漫/粘土/油画等）、美颜修图。当用户需要AI图片处理、人像编辑、背景替换、风格迁移、服装更换、脸部融合时使用此skill。支持用户通过描述性prompt（如"把衣服换成bikini"、"背景换成海边"、"转换成动漫风格"）进行图片编辑。

Security Scan

The skill largely matches an AI photo-editing purpose, but it embeds a hardcoded API key, fails to declare required credentials, and writes/opens files on the user's machine — these mismatches and secret-in-repo are concerning.

Purpose & Capability

The skill's name/description (AI photo editor) aligns with its code (image edit client, quota manager). However the package metadata declares no required credentials or env vars while the code expects an API key (scripts/config.py provides a default hardcoded key). That is an inconsistency: a production photo-editing integration would normally declare a required API key or ask the host to provide one, not ship a baked-in secret.

Instruction Scope

SKILL.md instructs normal photo-edit flows (URL/Base64/local file) and describes sending images to an external image-generation endpoint (api.vwu.ai). It also documents local behaviors: creating ~/.sih_ai/ usage files, auto-downloading outputs to ~/Desktop/Sih_Ai_Results/, and auto-opening Finder/Preview. Those file writes and GUI-open steps are outside pure API-calling functionality and are intrusive (and platform-specific to macOS). The instructions also route users to an external top-up URL that includes a user id parameter — expected for a billing flow, but notable for privacy.

Install Mechanism

No remote download/install spec is present (instruction-only install), which is lower risk. The package does include local Python scripts (client + quota manager). There is no installer that fetches arbitrary code at runtime from unknown hosts. Files present are local and readable for review.

Credentials

The skill metadata lists no required env vars/primary credential, but scripts/config.py contains a default API key (SIH_API_KEY = 'sk-w4Yf...'). Embedding a secret in the repo is poor practice and dangerous: that key may be valid, could be abused for API calls (costs/abuse), or leak activity tied to you. The skill will also transmit user images/prompts to a 3rd-party API (expected for image editing, but a privacy consideration).

Persistence & Privilege

The skill persists data under the user's home (~/.sih_ai) and creates/downloads results to ~/Desktop/Sih_Ai_Results/, and will attempt to open Finder/Preview on macOS. It does not request global "always" inclusion nor modify other skills. Persistent local files are expected for a quota/history feature, but users should be aware of where data is stored and that images (or prompt snippets) are written locally.

Scan Findings in Context

[hardcoded-api-key-in-repo] unexpected: scripts/config.py contains SIH_API_KEY default value 'sk-w4YfLvoX...'. A photo-editing client needs an API key, but shipping a default secret in the code and not declaring it in metadata is inconsistent and risks abuse/unauthorized charges.

[external-api-endpoint] expected: The code and SKILL.md use POST https://api.vwu.ai/v1/images/generations/ which is expected for an image-editing skill that calls a remote service. Users should know images and prompts are sent to this external endpoint.

[local-persistence-creation] expected: scripts/quota_checker.py creates ~/.sih_ai/ files and publish/README mention auto-downloading results to ~/Desktop/Sih_Ai_Results/. Persisting quota/history and result files is coherent with the described quota and auto-download features, but it's an important privacy/cleanup consideration.

What to consider before installing

Summary of what to consider before installing: - Do not assume the embedded API key is safe. The package includes a hardcoded API key in scripts/config.py. That key could be valid and might be used for API calls and billing. Ask the publisher to remove embedded keys and require a user-provided SIH_API_KEY via environment variable instead. If you proceed, rotate or revoke that key (and avoid using it for sensitive accounts). - Expect images and prompts to be sent to a third-party endpoint (api.vwu.ai). If you have privacy or legal concerns (face swaps, celebrity likenesses, or sensitive images), do not send those images. - The skill writes persistent files under ~/.sih_ai/ (user ID, usage_history.jsonl) and downloads outputs to ~/Desktop/Sih_Ai_Results/ and will attempt to open Finder/Preview on macOS. If you don't want that behavior, inspect/modify the scripts before use. - Metadata omits required credentials. Confirm with the author how API keys should be supplied and insist on explicit required env vars in the skill manifest. - If you don't trust the publisher or cannot verify the API key owner, do not install. If you need the functionality, request a version without embedded secrets and with clear consent for uploading images, or run the scripts in an isolated environment after reviewing/overriding the default key.

Sih.Ai - AI Photo Changer

让图片听你的话 —— 用自然语言描述，AI自动完成编辑

---

🚀 快速开始

用户上传图片 → 用一句话描述想要的改变 → AI处理 → 自动下载并预览

# 示例调用
from scripts.sih_api import SihClient

client = SihClient(api_key="your-api-key")
result = client.edit_image(
    image_url="https://example.com/photo.jpg",
    prompt="把衣服换成红色连衣裙，背景换成海边"
)
# 自动下载到 Desktop/Sih_Ai_Results/
# 自动在Finder中显示
# 自动在预览中打开

✨ 新功能：自动下载和预览

处理完成后，图片会自动：

1. 📥 下载到 ~/Desktop/Sih_Ai_Results/ 文件夹
2. 📂 在Finder中高亮显示
3. 👁️ 在预览app中打开

无需手动下载或二次处理，直接查看效果！

---

✨ 核心功能

人像编辑

• 换服装 - "把衣服换成bikini"、"穿上西装"、"换成汉服"
• 换发型 - "把头发剪短"、"染成金色"、"烫卷发"
• 换妆容 - "化个烟熏妆"、"加个口红"

背景处理

• 换背景 - "背景换成海边"、"背景改成城市夜景"、"背景换成咖啡厅"
• 移除背景 - "去掉背景"、"只保留人物"、"透明背景"
• 虚化背景 - "背景虚化"、"突出人物"

风格转换

• 艺术风格 - "转换成油画风格"、"水彩画效果"、"素描风格"
• 动漫/二次元 - "动漫化"、"二次元风格"、" manga style"
• 特效风格 - "粘土风"、"赛博朋克"、"复古胶片"

智能修图

• 人脸融合 - "把脸换成[明星名字]"、"换脸到这张图"
• 身材调整 - "瘦身"、"增高"、"调整身材比例"
• 美颜 - "磨皮"、"美白"、"去痘痘"

---

🔧 使用流程

1. 初始化客户端

from scripts.sih_api import SihClient
from scripts.quota_checker import QuotaChecker

# 使用Sih.Ai的API（需要用户提供API Key）
client = SihClient(api_key="sk-xxxxx")

# 或使用统一服务（需要用户ID和配额验证）
quota = QuotaChecker(user_id="user123")
if not quota.has_balance():
    print("余额不足，请充值：https://sih.ai/topup?user=user123")

2. 支持的输入格式

# URL格式
client.edit_image(
    image_url="https://example.com/photo.jpg",
    prompt="换服装成bikini"
)

# Base64格式
client.edit_image(
    image_base64="data:image/jpeg;base64,/9j/4AAQ...",
    prompt="背景换成海边"
)

# 本地文件（会自动转Base64）
client.edit_image(
    image_path="/path/to/photo.jpg",
    prompt="动漫化"
)

3. 高级参数

client.edit_image(
    image_url="...",
    prompt="换服装成bikini",
    model="sihai-image-27",  # 可选：模型选择
    size="1024x1024"  # 可选：输出尺寸
)

---

📝 Prompt最佳实践

✅ 好的Prompt

"把衣服换成红色bikini，背景换成海边沙滩"
"换装成白色婚纱，背景改成教堂"
"转换成动漫风格，背景改成樱花树下"
"脸换成Angelababy，保持原姿势"
"去掉背景，只保留人物"

❌ 避免的Prompt

"P一下"          # 太模糊，AI不知道具体要做什么
"好看一点"        # 主观描述不够具体
"修一修"          # 需要明确具体修改内容

💡 Prompt技巧

1. 具体描述 - "红色连衣裙" 比 "好看的裙子" 更好
2. 组合操作 - 可以同时要求换装+换背景+换风格
3. 保留元素 - "保持原姿势"、"保持面部表情"
4. 风格明确 - "日系动漫"、"美式漫画"、"3D渲染"

---

💰 配额与计费

用户自带API Key模式

client = SihClient(api_key="user_provided_key")
# 直接调用，计费由用户自己在Sih.Ai后台处理

统一服务模式（推荐）

from scripts.quota_checker import QuotaChecker

quota = QuotaChecker(user_id="user123")

# 检查余额
if not quota.check_balance():
    # 返回充值链接
    quota.show_topup_url()
    # https://sih.ai/topup?user=user123&return=xxx
else:
    # 扣除配额并处理
    quota.deduct(credits=10)
    result = client.edit_image(...)

计费建议

• 基础操作 - 5-10积分（换背景、简单换装）
• 复杂操作 - 15-20积分（换脸、风格转换）
• 批量处理 - 按图片数量累加

---

🎯 触发场景

优先使用此skill当用户提到：

直接触发

• "换服装"、"换背景"、"换脸"、"换发型"
• "P图"、"修图"、"美颜"、"精修"
• "动漫化"、"二次元"、"粘土风"、"风格转换"
• "AI处理图片"、"图片生成"、"图片编辑"

场景触发

• "这张照片帮我修一下"
• "能不能把背景换成xxx"
• "我想试试xxx风格的照片"
• "帮我P成动漫角色"
• "照片里的人能不能换成xxx"

关键词

人像：换装、换脸、换发型、化妆、美颜、瘦身、增高
背景：换背景、去背景、虚化、替换场景
风格：动漫、油画、水彩、素描、粘土、赛博朋克、复古
操作：修图、P图、美化、精修、处理、转换、生成

---

🛠️ 技术细节

API Endpoint

POST https://api.vwu.ai/v1/images/generations/

认证方式

Authorization: Bearer sk-xxxxx

请求参数

{
  "image": ["https://example.com/photo.jpg"],
  "prompt": "把衣服换成bikini",
  "model": "sihai-image-27"
}

响应格式

{
  "model": "sihai-image-27",
  "created": 1773386658,
  "data": [
    {
      "url": "https://...",
      "size": "2048x2048"
    }
  ],
  "usage": {
    "generated_images": 1,
    "output_tokens": 16384,
    "total_tokens": 16384
  }
}

---

📚 参考资料

• 完整API文档 - 见 references/api_guide.md
• Prompt示例 - 见 assets/examples/prompts.txt
• 配额验证API - 需要Sih.Ai提供余额查询endpoint

---

⚠️ 注意事项

1. 图片限制 - 确保图片URL可公网访问，或使用Base64格式
2. API Key安全 - 不要在代码中硬编码API Key
3. 配额管理 - 每次操作前验证余额，避免处理失败
4. 错误处理 - API失败时返回友好提示给用户

---

Sih.Ai官网： https://sih.ai
充值中心： https://sih.ai/topup
技术支持： [e-mail]