Poster Maker

AI 海报制作大师。三种模式：风格复刻、创意生成、直播预告海报。含产品原图还原方法论和生成脚本。Nano Banana Pro 生成，支持 Ofox/OpenRouter。

Security Scan

The skill appears to be what it says (an AI poster generator) but has several implementation and metadata inconsistencies around API keys and endpoint handling that you should review before installing.

Purpose & Capability

The skill's stated purpose (poster generation via Nano Banana Pro, supporting Ofox/OpenRouter) matches the included prompt templates and generation script. However the package metadata declares no required environment variables or credentials while the bundled script clearly needs an image API key. That mismatch between what is declared and what the code needs is inconsistent.

Instruction Scope

SKILL.md stays within poster-generation scope and instructs running the provided script; it does not instruct broad system access. But the runtime instructions claim automatic Ofox/OpenRouter detection while the script always posts to the Ofox API endpoint. The SKILL.md and script also use different names for the required key (the script reads OFOX_API_KEY or OPENROUTER_API_KEY but prints an error asking to set IMAGE_API_KEY), which is confusing and could cause accidental use of the wrong secret.

Install Mechanism

No install spec (instruction-only plus a small script). No downloads or archive extraction. This is lower-risk from an installation perspective.

Credentials

Metadata lists no required environment variables, but scripts expect OFOX_API_KEY or OPENROUTER_API_KEY; error messages reference a different name (IMAGE_API_KEY). Requiring an API key for an image-generation service is reasonable, but the unclear/mismatched env var names and the acceptance of multiple provider keys (without switching endpoints) are disproportionate and risky: a key intended for one service may be sent to a different endpoint.

Persistence & Privilege

The skill does not request persistent/always-on privilege, does not modify system-wide settings, and has no config path requirements. Autonomous invocation is allowed by default but is not combined with other high-risk flags.

What to consider before installing

This skill appears to do poster generation as described, but there are several sloppy/incoherent details you should resolve before use: - The script requires an API key (it reads OFOX_API_KEY or OPENROUTER_API_KEY) but the skill metadata claims no required env vars—ask the author to declare the needed env var(s) explicitly. - The script always POSTs to https://api.ofox.ai even if you supply an OPENROUTER_API_KEY; OpenRouter support is claimed but not implemented in the code. Do not assume a key for one service will be routed correctly. - Error messages reference IMAGE_API_KEY while code checks OFOX_API_KEY/OPENROUTER_API_KEY; this mismatch increases the chance you'll accidentally expose the wrong credential. - If you install/use this skill: provide a minimal-scope image API key, not a broad platform key or credential that grants other access. Prefer creating a dedicated API key with limited permissions and monitor usage/costs. - If you have sensitive secrets, test the script in an isolated environment and inspect network calls (or run it with a harmless test key) before using production keys. - Ask the author to fix metadata (declare required env vars), clarify OpenRouter support (implement switching endpoints or remove claim), and correct error messages. If these issues are unresolved, treat the skill as untrusted for production use.

AI 海报制作大师

三种模式，覆盖海报制作全场景。自带生成脚本，安装即用。

适用场景

电商产品海报 / 直播预告海报 / 营销活动海报 / 社交媒体素材 / 品牌宣传

前置：Brainstorming（可选）

需求不明确时先用 think-first skill 做一轮 brainstorming：产品是什么？目标人群？场景？有参考吗？根据回答进入对应模式。

三种模式

模式	输入	输出
1 风格复刻	参考海报 + 替换内容	同风格新海报
2 创意生成	产品图 + 卖点	原创海报
3 直播预告	品牌 + 产品 + 直播信息	直播预热海报

---

模式 1：风格复刻

Step 1：分析参考海报——用 7 维度（构图/色调/字体/装饰/排版/信息密度/气质）输出风格分析卡。详见 references/风格分析指南.md

Step 2：收集替换内容——产品图 + 主标题 + 卖点 + 品牌 + 尺寸

Step 3：生成 Prompt——风格分析 + 新内容合成。详见 references/Prompt模板.md

Step 4：生成 + 迭代——通常 3-7 轮

模式 2：创意生成

Step 1：理解产品——产品名/核心卖点/目标人群/品牌调性/参考方向

Step 2：AI 写文案——生成 2-3 套候选（主标题 + 副标题 + 卖点），用户选一套。详见 references/文案原则.md

Step 3：设计构图——产品居中型/场景代入型/信息密集型/极简留白型/对比展示型

Step 4-5：生成 Prompt → 生成 + 迭代

模式 3：直播预告海报

直播海报必备 10 个元素（缺一不可）：品牌名 / 直播日期时间 / "锁定直播间" / 平台标识 / 活动名称 / 品类标签 / 产品展示 / 产品卖点 / 底部活动栏 / 预约按钮

常见版式：竖版多品展示（2-4 个主推品 9:16）/ 横版单品主推（16:9）

风格推荐：深蓝星空+金丝绸（家纺/珠宝）/ 红金喜庆（大促）/ 清新渐变（食品/母婴）/ 黑金科技（3C/数码）

---

产品原图还原

AI 无法 100% 还原真实产品照片。三种应对策略：

策略 A 纯 AI（推荐）：用极详细的文字描述产品外观。还原度 70-85%，整体协调。

策略 B 模板+合成：AI 生成无产品模板 → 真实产品图抠图后合成。100% 还原但需 PS 技能。

策略 C 混合：AI 生成完整海报 → PS 只替换产品区域。兼顾协调和还原。

产品描述黄金模板：形状 + 颜色 hex + 纹理（类型/间距/凹陷深度）+ 材质 + 观察角度 + 特殊细节

迭代顺序：第 1 轮形状 → 第 2 轮颜色 → 第 3 轮纹理 → 第 4 轮质感

---

生成脚本

运行：

python3 scripts/generate_poster.py --prompt prompt.txt -o poster.png
python3 scripts/generate_poster.py --text "生成一张..." -o poster.png --size 1080x1080

自动检测 Ofox/OpenRouter，成本约 ¥0.1-0.3/张。

QA 迭代

通常 3-7 轮。每轮只改 2-3 个问题，不要推翻重来。

常见问题：中文乱码（加 CRITICAL）/ 产品颜色偏（用 hex）/ 产品太小（写"占 70%"）/ 缺直播元素（逐条检查 10 项清单）/ Prompt 泄漏（第一句改为描述句）

---

铁律

1. 文案用户确认后才生成
2. 产品描述要极其详细（hex + 纹理深度 + 材质 + 角度）
3. 中文必须清晰——加 CRITICAL 指令
4. 颜色用 hex——"浅绿色"不准确，#C8E6C5 才准确
5. 不编数据
6. 风格分析要具体——7 维度结构化
7. 直播海报别忘直播元素——10 项清单
8. 每轮迭代只改 2-3 个问题
9. 产品描述后期重点打磨
10. 构图先文字后图片