le-gpt-image v1.3：完整引入 GPT-Image2-Skill Reference Gallery（163 条精调 prompt / 31 类 / craft + 官方提示词指南，MIT，附 NOTICE 与原版 LICENSE，仅删预览图外链）；SKILL.md 工作流升级为 Gallery 优先（查库选案例做底改写，勿从零现写）；对照实测同主题海报 Gallery 框架质感显著优于即兴 prompt

skills/le-gpt-image/references/gallery/openai-image-2.5.md

@@ -0,0 +1,28 @@
+# GPT Image 2.5: optional reference index
+
+Keep Image 2's [gallery](gallery.md), [craft](craft.md) and
+[historical Cookbook](openai-cookbook.md) available; do not preload them for 2.5.
+A precise 2.5 brief can go straight to preflight and the CLI after model confirmation.
+
+## Choose only what is missing
+
+For guidance, open one relevant slice directly; reading this index is optional.
+
+| Need | Read |
+|---|---|
+| Photo, product scene or illustration | [Generation](openai-image-2.5-generation.md) |
+| Text, UI, diagrams or panel layout | [Layout and text](openai-image-2.5-layout-and-text.md) |
+| References, translation, masks or cutouts | [Editing](openai-image-2.5-editing.md) |
+| Migration, quality comparison or evaluation | [Migration notes](openai-image-2.5-migration.md) |
+| API parameters, errors or test status | [Model notes](models.md) |
+
+Translation starts with editing; a reference pose sheet may also need the panel
+section. Add a second slice only for that specific gap. [Community templates](templates-gpt-image-2.5.md)
+and older craft/gallery examples are optional inspiration, not a loading checklist.
+
+## Source boundaries
+
+Checked **2026-09-09**. These are locally organized official-source notes,
+not a verbatim 2.5 Cookbook or locally reproduced outputs. See
+[source details](openai-image-2.5-migration.md#source-boundaries) only when needed;
+[current validation status](models.md#validation-boundary) remains separate.