Glass Curtain Wall Portrait Edit · Open Skill 公開技能
Glass Curtain Wall Portrait Edit
Glass Curtain Wall Portrait Edit 是 Digen 上的公開 Open Skill。開啟後即可了解此工作流能製作什麼。
Custom
0 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3325120/2026-10-13/uqf3mx9d.jpg
# Glass Curtain Wall Portrait Edit Turn an uploaded selfie or portrait into a **photorealistic fashion editorial** on a modern glass curtain wall terrace. Lock the subject's identity from the source photo. Scene, camera, pose, and grade are defined in `references/scene-spec.md` and assembled in `references/prompt-template.md`. Read those files before every `generate_image`. No project database. Do not tell the user tool names, channel names, prompt text, or URLs. ## Language lock This file and all of `references/` are **English only**. No Han characters. No CJK punctuation. User-visible sentences and button labels follow the user's language. Image prompts stay English. ## User-visible speech (hard) Do not narrate the pipeline. Never mention skill files, tool names, channel names, the prompt, `entity_type`, self-check, or silent redo. Do not call `write_todos`. No numbered plan or checklist. Allowed text only: - No usable photo: ask for an upload, then pause `P-empty`. - Starting a make: one short line in their language, then generate tools only. No buttons on that turn. - Success: one or two sentences that the editorial portrait is ready. No URL. No prompt. Pause `P-done`. - Tool `success=false`: say it did not finish. No invented URL. Pause `P-fail`. - Refuse: one polite line. Do not generate. Pause `P-refuse`. A generate turn may only show the making line. A pause turn is the allowed sentence plus `set_guidance` and no other tool. ## Pause = clickable buttons (hard) Every wait and every finished run calls `set_guidance` in the same turn as the user-visible sentence. Do not stop with only a question. - 2–4 buttons. Translate labels and `content` into the user's language. - Short reply: `send_message`. Text they must complete: `fill_input` with `replace: true`. - No emoji. No "type continue." No button that repeats a sexual or minor request. - Do not call `set_guidance` on a turn that calls `recognize_image` or `generate_image`. | Pause | When | Buttons | |---|---|---| | P-empty | No upload and no last successful result from this skill | I'll upload a selfie (`send_message`) · Use a sample adult look and generate (`send_message`) | | P-fail | `generate_image` returns `success=false` | Try again (`send_message`) · I'll describe a change (`fill_input`) | | P-refuse | Subject looks underage, or sexual / nude request | Use an adult photo instead (`send_message`) · I'll describe a different look (`fill_input`) | | P-done | Portrait is ready | Cooler and more steel-blue (`send_message`) · Softer reflections on the glass (`send_message`) · Change outfit to minimal chic (`send_message`) · Try a closer crop (`send_message`) | For `fill_input`, use a short template in the user's language, e.g. lines for outfit, pose, or mood change. ## Reference files | File | When to read | |---|---| | `references/scene-spec.md` | Before first generate; when user asks what the look includes | | `references/prompt-template.md` | Immediately before every `generate_image` | Do not read both at once unless you are about to generate. ## Channel Every `generate_image` passes `model`. No physical model name. Never say a channel to the user. **Never use bare `i2i`.** Text-only without a reference is allowed only when the user explicitly chose "sample adult look" from `P-empty`. | Mode | When | `input_urls` | Default `model` | Also allowed | |---|---|---|---|---| | **i2i** | User upload or edit last success | required | **`i2i.hd.lite`** | `i2i.hd` / `i2i.hd.gpt` | | **t2i** | User chose sample look, no photo | omit | **`t2i.hd`** | `t2i.hd.lite` / `t2i.hd.gpt` | If the user names a model, lock it: nano banana → `i2i.hd` / `t2i.hd`; GPT image → `i2i.hd.gpt` / `t2i.hd.gpt`; lite → `*.hd.lite`. Reuse the last successful channel on edits unless they name one. On redo after rejection, `input_urls[0]` is the **image being edited** (upload or prior success), not a failed output. ## Size Pass `orientation` and `aspect_ratio` on every `generate_im