back
loading skill details...
Generate and adapt evidence-aware commerce, shopping, recommendation, and social content packs across platforms. Use when the user asks for product copy, 种草文...
---
name: content-generator
description: Generate and adapt evidence-aware commerce, shopping, recommendation, and social content packs across platforms. Use when the user asks for product copy, 种草文案, 小红书笔记, 抖音口播, 直播带货话术, 朋友圈推荐, 知乎评测, 电商详情页文案, ad copy variants, content repurposing, or safer marketing-claim review for ecommerce products.
---
# Content Generator
Create platform-ready commerce content without inventing unsupported claims.
## Workflow
1. Capture the product brief:
- product name, category, price or price range
- target audience and use scenario
- benefits, differentiators, proof points, limitations
- platform, tone, length, conversion goal
- forbidden claims or regulated categories
2. If the brief is thin, ask for the missing business-critical facts. If the user wants speed, continue with explicit assumptions and mark `待补充证据`.
3. Select the right output pack:
- Xiaohongshu: personal discovery note, practical list, comparison note
- Douyin: short-video hook, spoken script, captions, CTA
- Live selling: opening, demo flow, objections, close
- Moments: concise social recommendation
- Zhihu: rational evaluation and purchase advice
- Marketplace: title, bullets, detail-page sections
- Ads: A/B angle variants and CTA options
4. Draft using only user-provided facts or clearly labeled assumptions.
5. Run a quality pass before final output: evidence, platform fit, claim safety, specificity, and next data needed.
## Deterministic Pack Generator
When the user wants a structured first draft, batch variants, or repeatable output, use:
```bash
node scripts/generate_content_pack.js --input product.json --format markdown
```
The input JSON can include:
```json
{
"name": "HydraGlow Cream",
"category": "美妆",
"price": "299",
"audience": "commuters with dry skin",
"scenario": "winter office skincare",
"benefits": ["light texture", "long-lasting moisture"],
"evidence": ["user test: 8-hour office day"],
"limitations": ["not a medical treatment"],
"tone": "calm and trustworthy",
"platforms": ["xiaohongshu", "douyin", "marketplace"]
}
```
Read `references/platform-playbook.md` when platform details matter. Read `references/quality-checklist.md` before publishing, reviewing risky copy, or handling health, finance, legal, children, food, supplements, or high-priced products.
## Output Contract
For user-facing answers, prefer this shape:
- `Brief`: product, audience, goal, assumptions.
- `Content`: platform-specific drafts.
- `Quality Notes`: unsupported claims, risky phrases, missing evidence.
- `Next Iteration`: what data would improve conversion or trust.
Do not claim guaranteed outcomes, medical effects, investment returns, legal compliance, official endorsement, scarcity, or comparative superiority unless the user supplied reliable evidence.
don't have the plugin yet? install it then click "run inline in claude" again.