Iridescent Street Style Studio — Open Skill
Iridescent Street Style Studio
Iridescent Street Style Studio is a public Open Skill on Digen. Open it to see what this workflow creates.
TryOn
1 0 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4284865/2026-10-21/ts83fgdw.jpg
# Iridescent Street Style Studio
Turn a short written description into one photorealistic full-body studio fashion photograph with iridescent streetwear styling. There is no structured project state: the user's words plus the canvas assets are the whole context.
## When to use
- The user describes a model, an outfit, a color story, or a mood and expects a finished street-fashion photo (not a sketch, not an illustration).
- The user wants the same look regenerated with changes ("same outfit but silver", "make it a wider shot", "darker studio").
- The user pastes a style reference in words (era, brand vibe, fabric, silhouette) and asks for a picture.
## Steps
1. **Parse the request.** Extract from the user's text: who is photographed (build, gender, age impression, hair), the outfit and its silhouette, the iridescent / holographic materials and colors, accessories and footwear, pose and expression, background, lighting mood, and any framing request. If the user gave only a few keywords, fill the gaps with the defaults in the reference file — do not interview the user for every field. Ask at most one short clarifying question, and only when the subject or the outfit is genuinely missing.
2. **Read the style reference before generating.** Call `read_skill_file` with the skill name and `references/style-and-prompt.md`. Use its Style / Subject / Layout / Camera and light / Must keep / Must avoid sections and its prompt template to assemble the prompt — do not generate from the user's raw sentence.
3. **Assemble the prompt in English.** Fill every slot of the template with this request's specifics: a concrete subject, an outfit description naming the iridescent fabrics and cut, accessories, pose, background, lighting, palette and mood. Keep the photoreal anchors (full-body studio fashion photo, real skin texture, fabric sheen, lens and aperture). Append the avoid list from the reference (no text, no watermark, no logos, no illustration look). Never leave placeholder brackets in the final prompt.
4. **Generate.**
`generate_image(prompt=<assembled English prompt>, model='t2i', orientation='portrait', name=<short look title>, entity_type='characters')`
- `orientation='portrait'` is the default because the subject is a full-body standing figure; use `landscape` or `square` only when the user explicitly asks for that framing.
- Leave `resolution` unset (platform default) unless the user asks for higher clarity, then pass `'1080P'`; do not invent pixel sizes.
- `name` should read like a look name (e.g. "Iridescent Street Look 01").
- Do not pass `group_id` or `index`.
5. **Self-check against the reference.** Compare the result with the Must keep / Must avoid lists: the user's stated details are all present, the figure is a complete full-body frame, and the image reads as a photograph rather than a render or illustration. If the tool returns `success=false`, say so plainly and retry at most once with a simplified prompt (drop conflicting modifiers, keep subject + outfit + studio setup); never pretend it succeeded.
6. **Reply and suggest next steps.** Summarize the look in one or two natural sentences (outfit, materials, mood). Do not paste long URLs or raw prompts. If the user will likely continue, call `set_guidance` in the same turn with fitting options such as another colorway, a different pose, or a wider shot.
## Notes
- Preserve every detail the user stated — identity, outfit structure, colors, accessories — and never silently replace or drop them.
- Do not invent capabilities, brands, or settings the app does not describe; if a request falls outside iridescent street-fashion photography, say what this skill produces and offer the closest version.
- Refuse sexualized, violent, or otherwise disallowed content without calling `generate_image`.
- One generation per request; offer variations as a follow-up instead of generating a batch unprompted.