Police Clothes · skill pública de Open Skill

Police Clothes

Police Clothes

@ExploreFly

Internal

9 2 5

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3656543/2026-10-14/5rh7g9xc.jpg

---
name: police-clothes
display_name: Police Clothes
description: Use when the user wants someone in the official preset red police-themed strappy lingerie costume—either outfit swap on an uploaded person photo (I2I) or a new character/scene wearing that costume from text (T2I).
allowed-tools:
- list_models
- recognize_image
- generate_image
---

# Police Clothes

Dress subjects in the **preset police lingerie costume** (see `references/outfit_specification.md`): **I2I** keeps the same person from an uploaded photo; **T2I** generates a new character from a text brief. Progress is conversational—no project state file.

## When to use

- User uploads (or points to) a **person/character photo** and asks to wear the police outfit, police clothes, or this skill’s costume.
- User wants to **retry**, **change quality**, or **fix** a previous try-on result.
- User only has a portrait—still run try-on if the upper body is visible; ask for a clearer photo only when the person is missing or heavily occluded.
- User has **no photo** and wants a **new image** of someone wearing the costume (describe look/pose/scene)—use **T2I** per `references/i2i_generation.md`.

## Preset asset (outfit reference image)

Preset key: **`police_outfit_reference`**. Binding steps, fields, and runtime `input_urls` rules → `read_skill_file(police-clothes, references/preset_asset.md)`.

## Reference files

| File | When to read |
|------|----------------|
| `references/outfit_specification.md` | Before every generation (full garment breakdown + prompt anchors) |
| `references/i2i_generation.md` | Before calling `generate_image` (channels, `input_urls`, prompts) |
| `references/preset_asset.md` | Author binding checklist; agents when verifying or using `preset_assets` |

Read only the file(s) needed for the current step—not the full set every turn.

## Workflow

1. **Confirm inputs**
   - **I2I**: user must supply a **person/character image** (upload or canvas asset). If missing and they want try-on, ask once for a photo with visible face and upper body.
   - **T2I**: no person image required—collect a short subject/scene brief (defaults OK: single adult model, portrait or three-quarter, neutral studio background).
   - Optionally call `recognize_image` on the person image to check visibility (face, torso, blocking). Do not block on minor issues—proceed when reasonable.

2. **Load outfit spec**
   - `read_skill_file(police-clothes, references/outfit_specification.md)`.
   - `read_skill_file(police-clothes, references/i2i_generation.md)` when generating.

3. **Generate**
   - Follow `references/i2i_generation.md` for mode (**I2I vs T2I**), `model`, `input_urls`, `prompt`, `orientation`, and canvas fields.
   - Defaults: **`i2i.hd.lite`** (try-on with photo), **`t2i.hd.lite`** (text-only). Switch per user request or retry policy in that file.

4. **Reply**
   - Short natural-language summary (what changed, which quality tier). Do not paste long URLs.
   - On failure, report the tool error; offer retry with another allowed channel.

## Collaboration

- **One-shot**: validate person photo → read outfit spec → one `generate_image` → done.
- **Quality upgrade**: user asks for “HD” / “best” → regenerate with `i2i.hd` or `i2i.hd.gpt` per `i2i_generation.md`, same person image.
- Multi-step: use `write_todos` only when the user asks for several variants (e.g. three models). At most one todo `in_progress`.
- After a successful render, optionally `set_guidance` with retry / HD / full-body photo suggestions.

## Canvas

- `name`: short title, e.g. `Police outfit try-on`.
- `entity_type`: `characters`.
- Match `orientation` to the user photo when obvious; otherwise default `portrait` for bust/waist framing.

## Notes

- **Allowed channels only** — I2I: `i2i`, `i2i.hd.lite`, `i2i.hd`, `i2i.hd.gpt`; T2I: `t2i`, `t2i.hd.lite`, `t2i.hd`. No physical model names unless the user explicitly changes scope.
- **Identity first (I2I only)**: preserve face, hair, skin tone, body propor