Objects After Dark · Open Skill 公開技能
Objects After Dark
Objects After Dark 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/4173183/2026-10-15/c5m6s9uz.jpg
# Objects After Dark **Product (read first):** *When everyday things wake up.* Deliver **exactly one shown image** — a **cinematic Objects After Dark key-art still**, not a mild retouch and **not** video. The picture is a **bold, exaggerated, visually rich** key-art poster: **many** everyday objects awake at once with **big eyes and smiles**, **dynamic mischievous poses**, and **dense golden night magic** (premium 3D-photoreal animated-film look). **Do not** aim to preserve the user's photo layout or camera — uploads only inspire **room type** and **prop ideas** via `recognize_image`, then **fully reimagine** with **`t2i.hd`**. **Image-only (hard):** **`generate_image` only**. Never `generate_video`, `video_agent`, or `mix_clips`. - **Upload path:** `recognize_image` → **`t2i.hd`** (no `input_urls`) + `references/poster_image_prompt.md`. - **Direct generate:** **`t2i.hd`** + `references/scene_still_spec.md`. - **Refine last poster:** **`i2i.hd`** on the last **Objects After Dark** image only — not on raw upload. **Language lock:** This file and all of `references/` are **English only**. Reply in the **user's language**. Every `generate_image` `prompt` is **English**. Do not tell the user tool names, channels, or full prompts. ## User-visible speech (hard) - Need input → **P-empty** / **P-brief** - Starting generate: one short line, then tools only — **no** `set_guidance` that turn - Success: the **picture** is ready → **P-done** - Failure / refuse → **P-fail** / **P-refuse** ## Pause = clickable buttons (hard) When stopping without a new shown **image**, or on success: 1. One or two short sentences (user language). 2. **`set_guidance`** same turn, **≥2** buttons (`text` + `action`). **No tools after.** Do **not** call `set_guidance` with `generate_image` or `recognize_image` in the same turn. Before pause: `read_skill_file` → `references/guidance-buttons.md`. | Pause | When | |---|---| | P-empty | No upload, no brief, no direct-generate tap | | P-brief | Vague text brief | | P-refuse | Policy block | | P-fail | `generate_image` failed after retry | | P-done | Objects After Dark **image** ready | ## Reference files | File | When | |---|---| | `references/poster_image_prompt.md` | Before **every** `generate_image` (upload I2I and refine) | | `references/scene_still_spec.md` | Before **T2I** direct generate | | `references/generation.md` | Channels, HD, retries | | `references/guidance-buttons.md` | Every `set_guidance` pause | ## Steps 1. **Intake** - Upload → **`recognize_image`** then **`t2i.hd`** reimagined poster (see `references/generation.md`). - Direct generate / button → **`t2i.hd`** + scene spec. - Refine last **Objects After Dark** image → **`i2i.hd`** on that asset only. - More chaos / stronger magic → refine or new **`t2i.hd`** with exaggerated beats in prompt. - Empty → **P-empty**. 2. **`recognize_image`** (upload only) — room type, prop list, hero hint, policy. **Do not** instruct preservation of layout or likeness in later prompts. 3. **`read_skill_file`** as needed from table above. 4. **Size:** default **landscape** poster framing unless user asks portrait/square. 5. **`generate_image`** (one shown result) - **Create:** always **`t2i.hd`**. **Refine prior poster:** **`i2i.hd`**. Never `.lite` unless user asks faster. - Prompt must pass **Poster check** (exaggerated, many faces, dense magic, **not** upload-like). - If tame or upload-like → **Poster redo** retry. - `name`: `Objects After Dark` - `entity_type`: `props` 6. Reply + **`set_guidance`**. ## Collaboration - **Direct generate is first-class.** - One image per request unless user asks variants. - Never invent URLs. ## Content policy Magical, charming, family-friendly. Refuse sexualized minors, explicit nudity, graphic gore, hate imagery. ## Author publish checklist (not for end users) 1. `skill validate` 2. `skill publish --activate`