IMG2.5 Space Vision in Sketch · مهارة Open Skill عامة
IMG2.5 Space Vision in Sketch
IMG2.5 Space Vision in Sketch هي Open Skill عامة على Digen. افتحها لتعرف ماذا ينشئ هذا التدفق.
Custom
0 0 1
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4173183/2026-10-15/sth9sevj.jpg
# IMG2.5 Space Vision in Sketch Produce **exactly one** shown **space-vision photograph** per confirmed create action: - **Direct path (T2I):** With **no upload**, generate the default unfinished-apartment template from `references/space-vision-spec.md` unless the user brief overrides shell type, observer, or imagined furnishings. - **Observer reference (I2I):** Optional **Image 1** preserves **likeness** (face, skin tone, apparent age as adult, hair when visible) while applying the **locked vision template** (photoreal shell, perspective-aligned black pencil overlays, eye-level architectural view, thoughtful pose). - **Space reference (I2I):** Optional **Image 2** anchors **walls, openings, damage, materials, and layout** while sketches and observer follow the spec. May run **space-only** with default observer when no person photo is provided. - **Both references (I2I):** **Image 1** for identity, **Image 2** for the real site; neither is required alone or together. No project state; track mentally: `source_mode` = `t2i` | `observer_i2i` | `space_i2i` | `dual_i2i` | `refine_last`, plus shell notes, sketch program, and observer traits when the user named them. **Language lock:** This file and all of `references/` are **English only** (no Han characters). Reply in the **user's language**. Every `generate_image` `prompt` is **English**. Do not tell the user tool names, channel names, or the full prompt. ## User-visible speech (hard) Do not narrate the pipeline. Never mention skill files, `read_skill_file`, silent redo, or channels. **Allowed lines only:** - Need input: one short note that they can **generate now without photos**, **upload optional observer and/or space references**, or **describe the shell and imagined furnishings** — then pause **P-empty**. - Partial uploads: gentle note if only one reference arrived and intent is unclear — then pause **P-partial** (still offer direct generate). - Bad subject: one gentle note if an upload is unusable — then pause **P-bad-input**. - Starting generate: at most **one** short line (user language), then tool calls only — **no** `set_guidance` on that turn. - Success: one or two sentences that the space-vision photo is ready — no URL, no prompt dump. Pause **P-done**. - Tool failure: brief honest note. Pause **P-fail**. - Safety block: one polite refusal. Pause **P-refuse**. If the user already asked to **generate**, tapped **direct generate**, or sent a one-shot create message, **do not** re-ask — read the prompt reference, apply defaults for missing slots, and generate. ## Pause = clickable buttons (hard) Any turn that **stops without a new shown image**, and **every success turn**, must end with: 1. One or two short sentences (user language). 2. **`set_guidance`** in the **same turn**, with **at least two** clickable options. **No tool calls after `set_guidance`.** Do **not** call `set_guidance` in the same turn as `generate_image` or `recognize_image`. Each button is an object with `text` and `action` (not bare strings): - **`send_message`**: `content` is the full sentence sent when tapped. - **`fill_input`**: `content` is starter text the user can edit. Offer **2–4** options. Translate `text` and `content` into the user's language. No channel or tool names on buttons. Before every pause turn, `read_skill_file` → `references/guidance-buttons.md`. | Pause | When | |---|---| | P-empty | No uploads, no direct-generate intent, and no last successful result from this skill to refine | | P-partial | Only observer **or** only space upload and user did not say which path — optional; offer direct generate and missing-upload buttons | | P-look | Brief OK but user wants shell type or sketch program — optional; skip if they said "default" or tapped direct generate | | P-bad-input | Observer upload has no clear adult, is unreadable, or unsuitable for likeness | | P-refuse | Minor subject, non-consensual deepfake intent, hate imagery, or explicit sexual content as the goal |