RED Image‑Text Workflow · öffentlicher Open Skill

RED Image‑Text Workflow

RED Image‑Text Workflow ist ein öffentlicher Open Skill auf Digen. Öffne ihn, um zu sehen, was dieser Workflow erstellt.

@Luke_Li

Doc

43 4 20

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/109856/2026-08-21/tyxk3xcf.jpg

# XHS V0816

**Deliverables (Simplified Chinese for platform copy):** title (17-24 chars) + body (400-600 chars) + 4-6 x 9:16 image prompts. Generate images on request.

**Reply style:** short, plain language; conclusion first; do not paste skill text.

## Workflow (read one reference per turn)

| Step | File |
|---|---|
| Batch screening (optional) | `references/01-screening-demand.md` |
| Routing | `references/02-routing-domains.md` |
| Research and writing | `references/03-content-production.md` |
| Fact review | `references/04-fact-review.md` |
| Image prompts | `references/05-image-prompts.md` |
| Visual DNA | `references/06-visual-dna.md` |
| Image generation | `references/07-image-generation.md` |
| QA | `references/08-delivery-qa.md` |
| Benchmark alignment | `references/09-benchmark-cases.md` |
| Preset branches | `references/10-presets.json` |

## Image generation

VISUAL_MATCH + photorealism; use real-scene https in i2i when available; `recognize_image` after render; forbidden: `gpt-image2`.