Yoga Pose Nine-Grid · skill pública de Open Skill

Yoga Pose Nine-Grid

Yoga Pose Nine-Grid

@ExploreFly

Health

1 1 0

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3656543/2026-10-14/x2tg75mp.jpg

# Yoga Pose Nine-Grid

Turn the user’s **uploaded person photo** into **one square 3×3 showcase image**: nine equal panels, each panel a **different classic yoga asana**, **same person** as the reference, modest athletic wear, clean readable form. Primary output is **one contact sheet** on canvas (`Yoga 3×3 grid`). Optional follow-up: **nine separate tiles** via `split_image` (hidden from canvas by default; tiles shown per step below).

No structured project state. Progress is conversation + canvas assets.

## Handbook language

This file and all `references/` are **English only**. Reply to the user in **their language** (default Chinese if unclear). Image **`prompt` is always English**.

## Reference files (progressive disclosure)

Do **not** read everything at once.

| File | When |
|---|---|
| `references/yoga-poses-nine-grid-spec.md` | Before the first `generate_image`; includes the **nine asanas**, per-pose visual cues, and the **verbatim contact-sheet prompt skeleton** |
| `references/grid-split-and-qc.md` | Optional split into nine images; silent QC checklist |

Do not tell the user you are reading these files.

## When to use

- User uploaded a photo and wants a **九宫格 / 3×3 / nine-panel** yoga pose sheet.
- User wants **different yoga poses** in one composed image while **keeping the same face and body identity**.
- User asks to **regenerate**, switch **HD / fast / GPT** quality, change **mat color / studio background**, or **export nine single pose images**.

Do **not** use for: minors, nudity or explicit sexual content, fetish or overly revealing “yoga” framing, violence, or hate. Refuse without calling generate.

## Image-to-image channels (required)

Every `generate_image` that uses the user upload must set **`model`**. **Only these three channels** — never bare `i2i`, never physical model names, never `t2i*` when an upload is present.

| Channel | When |
|---|---|
| **`i2i.hd.gpt`** | **Default** for the first full 3×3 grid from upload |
| **`i2i.hd.lite`** | User asks faster / lighter / draft / cheaper |
| **`i2i.hd`** | User asks **HD**, **max detail**, **best quality**, or sharper limb/finger definition |

| Call | `input_urls` |
|---|---|
| **First grid from upload** | `[user_upload]`; if identity drifts on QC, retry once with `[user_upload, user_upload]` |
| **Regenerate whole grid** | `[user_upload]` or `[user_upload, last_grid]` if user wants pose layout similar but fixed |
| **Regenerate after QC fail** | Same channel unless user changed tier; prefer `[user_upload]` |

## Size and canvas fields

- **`orientation='square'`** for the 3×3 contact sheet (required unless user explicitly wants a non-square sheet — then explain tradeoffs and still use equal cells).
- Default **`resolution='1080P'`**; user wants sharper → `'2K'` or `'4K'` (platform may clamp).
- **`entity_type`**: `yoga_pose_nine_grid`
- **`name`**: `Yoga 3×3 grid` (or user-language short title).
- **Contact sheet**: `add_to_canvas=true`.
- **Optional split pass**: `split_image` with `add_to_canvas=true` on tiles only; contact sheet may stay on canvas from step 4.

## Steps

1. **Intake**
   - Require an upload unless the user already attached one in-thread.
   - Confirm: keep **original visible outfit** vs switch to **neutral yoga set** (leggings + fitted top); default **yoga athletic wear** if upload is street clothes and user did not insist on keeping outfit.
   - Note quality tier → channel table above.
   - If user only asks what the nine poses are, read the spec reference and list names in user language — **do not generate** unless they want the image.

2. **Recognize upload (required)**
   - `recognize_image` with prompt: `For a 3×3 yoga pose grid: (1) apparent age — adult only, refuse if minor or ambiguous youth; (2) face — eyes, brows, nose, lips, outline, skin tone; (3) hair color, length, style; (4) visible outfit and footwear; (5) body framing (head-to-toe vs waist-up) and current pose; (6) lighting direction. Internal use only.`

3. *