Mode Prompts

# Mode Prompts · 三模式 prompt 分流

> Three pre-configured "personas" the AI adopts based on the writer's role. The mode shapes voice, source-attribution patterns, what counts as a "good" article, and what counts as a violation.

## Why three modes?

Industry research, corporate communications, and viral content are **three different crafts** with three different value systems. Asking AI to "write an article" without specifying which craft produces an averaged-out result that satisfies no one.

Editorial AI's three modes encode the distinct value systems explicitly:

| Mode | What "good" means | Failure mode it prevents |
|---|---|---|
| 🎯 **KOL Mode** | Reader wants to share / save / screenshot | Boring corporate-speak, no hook, no punchline |
| 📰 **Editorial Mode** | Reader (industry insider) wants to cite / commission you | Slick marketing-AI tone that breaks journalism norms |
| 🏢 **Corporate Mode** | Reader (investor / client / employee) trusts the company more | Self-promoting fluff that triggers eye-rolls in pros |

## 🎯 KOL Mode — for viral / marketing content

**No additional prompt addendum.** This is the default mode. Uses only the base skeleton + anti-fabrication + reflection rules.

Audience: social media creators, marketing teams, viral content, growth marketers.

Outputs: WeChat OA articles, Douyin scripts, Xiaohongshu notes, Twitter threads.

Scoring: 7 dimensions (hook / punchline / contrast / detail / rhythm / ending / solution).

## 📰 Editorial Mode — for industry research / journalism

**Prompt addendum to inject after the skeleton block:**