📋 Overview
Based on the Redfox Data official account viral radar, search current popular viral articles by keyword, extract traffic secrets and key creative points, and generate complete articles ready for direct publication.
Data Scope: Official accounts across the entire web, with viral articles continuously collected daily | Core Process: Query viral articles → Analyze patterns → Generate copy
🔑 Authentication
Get API Key
Please go to Redfox hub to get the API KEY
Configure API Key
Option 1: Using OpenClaw as an example, add REDFOX_API_KEY to ~/.openclaw/openclaw.json. Some of the content is as follows:
{ "env": { "REDFOX_API_KEY": "ak_xxxx..." } }
Option 2: Terminal configuration
export REDFOX_API_KEY="ak_xxxx..."
🔄 Workflow
Step 0: Intent recognition and keyword extraction
First understand the user's real needs, then extract search keywords:
1. Understand user intent
- Core demand: What type of content do they want to write, and is there a clear theme and argument
- Target audience: Who is it written for
- Creative purpose: Share experience, teach others, or express an opinion
2. Determine the type of writing subject
| Type | Basis for judgment | Keyword strategy |
|---|---|---|
| Specific entity | The user's description contains clearly identifiable object names (products, tools, books, etc.) | Intent keywords + subject keywords |
| General topic | Broad-domain topic (workplace, relationships, health, personal finance, etc.) | Core keywords + scenario keywords + expanded keywords |
3. Extract keywords
General topic: Core keywords + scenario keywords + expanded keywords
Specific entity(subject keywords must also be extracted):
| Dimension | Description | Example (gzh-copywriter) |
|---|---|---|
| Category term | What category it belongs to | AI skills, productivity tools |
| Positioning term | Core selling point keywords | Viral article generation, data-driven |
| User term | The terms target users use when searching | Official account magic tool, self-media magic tool |
Rule: subject keywords must be no fewer than 2, with priority no lower than intent keywords. Final search terms = intent keywords + subject keywords (merged and deduplicated). At least 2 different keywords must be extracted.: Subject keywords must be no fewer than 2, with priority no lower than intent keywords. Final search terms = intent keywords + subject keywords (merged and deduplicated). At least 2 different keywords must be extracted.
4. Examples
General topic: The user says "Help me write an article about workplace job-hopping" → Search terms: workplace, job-hopping, resignation, salary increase, career development
Specific entity: User says "help me write a WeChat official account article recommending gzh-copywriter" → Search terms: WeChat official account, viral articles, AI skills, productivity tools, viral content generation, data-driven, WeChat official account tool
Step 0.5: Differentiation advantage breakdown [executed only when recommending a specific entity]
Pure topic-based writing skips this step.
1. Information collection (try in order of priority, stop once sufficient)
| Priority | Source | Method of operation |
|---|---|---|
| First | Information already included in the user's description | Extract functions, features, advantages |
| Second | Readable structured documents | Read product documents, skill descriptions, etc. |
| Third | Online search | Search official introductions, user reviews, competitor comparisons |
2. Breakdown model (four dimensions)
| Dimension | Breakdown question |
|---|---|
| Input difference | What unique input sources does it rely on? |
| Process difference | How does its way of working differ from similar products? |
| Output difference | How do its results differ from similar products? |
| Positioning difference | What audience/scenarios does it target that competitors do not cover? |
3. Output format
Output 3-5 core differentiation advantages, each including:
优势描述：[一句话]
与竞品对比差异：[说明不同之处]
4. Downstream references
These 3-5 differentiation advantages must be mandatorily referenced in subsequent steps:
- Step 2: Compare the differentiation advantages with viral content patterns to find points of alignment
- Step 4: The article must be developed around the differentiation advantages as the "anchor"
Step 1: Query viral articles
python3 scripts/fetch_gzh_trends.py \
--keyword "<提炼出的关键词>" \
--max-items 10
Parameter description：
--keyword: Search keywords (separate multiple with English commas, maximum 5, total length ≤ 200 characters); if Step 0.5 was executed, must include the subject keyword--max-items: Number of articles to display, default 10--start-date: Optional, start date, format yyyy-MM-dd (maximum is the last 30 days). If not specified, defaults to the last 7 days; if data is insufficient, automatically extends to the last 30 days
Execution requirements: After executing the script, analyze directly based on the output data; it is forbidden to show raw viral data to the user, and forbidden to send only a file path or a brief prompt such as "saved".
Step 2: Analyze viral patterns
Based on the script output, extract the following patterns:
Title patterns: Common features of high-engagement titles (use of numbers, emotional words, audience labels), title length and structural patterns
Content patterns: Topic directions and entry angles of popular articles, content structure and paragraph distribution characteristics, interaction guidance methods
Differentiation advantage comparison(if Step 0.5 was executed): Compare the differentiation advantages with viral content patterns to find points of alignment
Data analysis dimensions：
| Metric | Analysis points |
|---|---|
| Read count | Content reach and appeal |
| Wow count | Content depth and bookmark value |
| Like count | Degree of emotional resonance |
| Comment count | Topic relevance and interaction guidance effectiveness |
Step 3: Guide the user to upload personal style references
Before generating the copy,mustproactively ask the user whether they will provide personal style references:
Want me to understand your style better? Just send me a few of your usual notes, diary entries, or essays
- If the user provides them: analyze their writing style (word choice habits, tone characteristics, expression methods) and integrate it into the copy
- If the user does not provide them: directly generate the copy based on viral content patterns
Step 4: Generate the copy
Must create strictly according to the viral content pattern analysis results.
Creation preparation
- Determine the core viewpoint(Required): Extract one core viewpoint from the user's intent, viral data, and the latest information, and develop the entire article around this viewpoint. The viewpoint must be novel, controversial, or counterintuitive. If Step 0.5 was executed, the core viewpoint should revolve around differentiated advantages
- Review viral content patterns(Required): Extracted high-frequency keywords (must be integrated into the main text), popular title patterns (must be followed), viral structure characteristics (must be referenced), popular content style (must be maintained)
- Integrate differentiated advantages(If Step 0.5 was executed): Use 3-5 differentiated advantages as the article's "anchors", with each advantage corresponding to a core argument or paragraph
- Search for the latest information(Optional): If timely material needs to be supplemented, you may use
web_search. Note: the latest information is only supplementary material; viral content patterns must come from script data
✍️ WeChat Official Account writing guidelines
Writing principles: unique entry point, simple and blunt title that directly hits pain points, counterintuitive viewpoints, emotional resonance, integration of real experiences, relevance to readers, practical value, interesting writing style, create a sense of urgency that there will be a loss if not read, distinctive
Word count requirements: main text around 1,500 characters (opening 200-300 characters, core 800-1,000 characters, ending 200-300 characters)
Style characteristics: clear attitude, frequent memorable lines, strong contrasts, directly hits pain points, tight rhythm, provokes thought
Title creation：
- Must reference the title structure that appears most frequently in viral data
- Prioritize numeric titles; interrogative/exclamatory sentences may be used
- Title must not exceed 20 characters (including punctuation)
Main text creation：
- Must integrate high-frequency keywords from viral data
- Must reference the opening "hook" techniques of viral content
- Must maintain the writing style of viral content
- Add interactive engagement wording at the end (e.g., "What do you all think?")
- Formatting: good paragraphing, embellish with a small number of Emoji (1-2 per paragraph), mainly short sentences
- Add 5-10 related tags at the end
Step 5: Self-check before output [Required]
Check item by item whether the output is complete:
- Recommended titles: 3-6
- Main text content: complete and publishable
- Core viewpoint: clearly stated
- Recommended tags: 5-10
- Viral formula sources - viral content patterns: 6 dimensions: title/opening/structure/language/interaction/high-frequency words
- Viral formula sources - reference articles: 2-3, including title, link, author, interaction data
- Differentiated advantages (if Step 0.5): 3-5, including advantage descriptions and differences compared with competitors
- Ask the user to upload a personal style reference
📊 Output format
Output plain copy content, including:
### 推荐标题
1. [标题1]
2. [标题2]
3. [标题3]
### 正文内容
[完整可发布的正文]
**核心观点**：[一句话概括]
### 推荐标签
#标签1 #标签2 #标签3 #标签4 #标签5
### 差异化优势（仅推荐具体实体时输出）
**优势1**：
- 优势描述：[一句话]
- 与竞品对比差异：[说明]
**优势2**：
- 优势描述：[一句话]
- 与竞品对比差异：[说明]
### 爆款公式来源
**数据来源说明**：以下爆款规律和参考文章均来自脚本获取的真实爆款数据，非网络搜索结果。
**参考的爆款规律**：
1. **标题规律**：[具体描述]
2. **开头规律**：[具体描述]
3. **结构规律**：[具体描述]
4. **语言风格**：[具体描述]
5. **互动引导**：[具体描述]
6. **高频关键词**：[3-5个高频词及频次]
**参考的爆款文章**：
1. [文章标题](链接) - 作者：公众号名称 - 阅读数：X / 在看：X / 点赞：X / 评论：X
2. [文章标题](链接) - 作者：公众号名称 - 阅读数：X / 在看：X / 点赞：X / 评论：X
Prohibited output: card design or images, script execution commands, file paths or download links, displaying data only without generating copy
⚠️ Notes
- Strictly analyze based on the real viral data queried
- The generated copy must follow the extracted viral patterns
- Do not add external contact information or violating information to the copy
- The generated copy must comply with WeChat Official Account community guidelines