neta-suggest

Neta Suggest Skill

Instructions

1. When handling research tasks such as "just help me find some themes," "help me see what's trending now," or "filter content by a certain topic/category," you can refer to the following recommended process:
2. Recommended standard process:Browse categories → discover tags → validate category path → get recommended content(The "Progressive Exploration Process" section of this document already provides detailed commands).
3. Before content creation, you can use this skill to first complete research at the theme/tag/category level, then hand it over to neta-creative for specific creation.
4. When the user needs to like/interact with a specific work, you should switch to neta-community。

Core Skills

1. suggest_keywords - Get search keyword suggestions

Provides popular search term suggestions based on the input prefix, helping users discover content directions they are interested in.

npx -y @talesofai/neta-skills@latest suggest_keywords --prefix "游戏" --size 20

Parameter description:

• --prefix: keyword prefix (required)
• --size: number of results to return, recommended 10-20 (optional, default value depends on system configuration)

Applicable scenarios:

• When the user only has a vague idea
• Explore popular topics and trends
• Prepare for subsequent precise filtering

2. suggest_tags - Get related tag suggestions

Recommends related tax tags based on a complete keyword.

npx -y @talesofai/neta-skills@latest suggest_tags --keyword "角色塑造" --size 15

Parameter description:

• --keyword: complete keyword (required)
• --size: number of results to return, recommended 10-20 (optional)

Applicable scenarios:

• Understand related tags for a certain topic
• Discover popular tags in vertical domains
• Choose suitable topic tags for content creation

3. suggest_categories - Get category hierarchy navigation

Provides navigation suggestions for a three-level category system, supporting level-by-level in-depth exploration.

# 获取一级分类（顶层分类）
npx -y @talesofai/neta-skills@latest suggest_categories --level 1

# 获取二级分类（需要父级路径）
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "衍生创作类"

# 获取三级分类（最细粒度）
npx -y @talesofai/neta-skills@latest suggest_categories --level 3 --parent_path "衍生创作类>同人二创"

Parameter description:

• --level: category level (1/2/3) (required)
• --parent_path: parent category path, required when level>1 (optional)

Category system:

一级分类（Level 1）
├─ 衍生创作类
│  ├─ 同人二创（Level 2）
│  │  ├─ 崩坏星穹铁道（Level 3）
│  │  └─ 原神（Level 3）
│  └─ 数字艺术
├─ 生活方式
└─ ...

Applicable scenarios:

• Systematically understand platform content categories
• Find content in specific domains by following the trail
• Validate the effectiveness of category paths

4. validate_tax_path - Validate category path

Checks whether a category path is valid, avoiding the use of nonexistent categories.

npx -y @talesofai/neta-skills@latest validate_tax_path --tax_path "衍生创作类>热门 IP>崩坏星穹铁道"

Parameter description:

• --tax_path: complete category path (required)

Applicable scenarios:

• Validate before using a category path
• Ensure the accuracy of the category path

5. suggest_content - Intelligent content flow engine

A powerful content recommendation tool that supports three modes: recommendation, search, and precise filtering.

# 模式 1：推荐模式（广泛探索）
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 0 \
  --page_size 20 \
  --scene agent_intent \
  --intent recommend

# 模式 2：搜索模式（关键词搜索）
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 0 \
  --page_size 20 \
  --scene agent_intent \
  --intent search \
  --search_keywords "角色,创意"

# 模式 3：精确模式（分类筛选）
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 0 \
  --page_size 20 \
  --scene agent_intent \
  --intent exact \
  --tax_paths "衍生创作类>同人二创"

# 组合使用（多条件筛选）
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 0 \
  --page_size 20 \
  --scene agent_intent \
  --intent search \
  --search_keywords "AI,绘画" \
  --tax_paths "数字艺术>概念设计" \
  --exclude_keywords "测试,废弃"

Parameter description:

• --page_index: page number, starting from 0 (default 0)
• --page_size: number per page, range 1-40 (default 20)
• --scene: scenario identifier (default "agent_intent")
• --biz_trace_id: session tracking ID (optional)
• --intent: intent type recommend | search | exact(default recommend)
• --search_keywords: Search keywords, separate multiple ones with English commas (optional)
• --tax_paths: Category path, separate multiple ones with English commas (optional)
• --tax_primaries: First-level category, separate multiple ones with English commas (optional)
• --tax_secondaries: Second-level category, separate multiple ones with English commas (optional)
• --tax_tertiaries: Third-level category, separate multiple ones with English commas (optional)
• --exclude_keywords: Exclude keywords, separate multiple ones with English commas (optional)
• --exclude_tax_paths: Exclude category paths, separate multiple ones with English commas (optional)

Progressive Exploration Process

Standard Exploration Path (from broad to narrow)

graph LR
    A[浏览分类] --> B[发现标签]
    B --> C[验证路径]
    C --> D[获取内容]

Step 1: Browse the Category System

First, understand the overall content category structure:

# 查看所有一级分类
npx -y @talesofai/neta-skills@latest suggest_categories --level 1
# 输出示例：["衍生创作类", "数字艺术", "生活方式"]

# 对感兴趣的分类深入查看
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "衍生创作类"
# 输出示例：["同人二创", "原创故事", "互动小说"]

Step 2: Discover Related Tags

Discover popular tags based on categories or keywords:

# 基于关键词找标签
npx -y @talesofai/neta-skills@latest suggest_tags --keyword "同人二创" --size 15
# 输出示例：["崩坏星穹铁道", "原神", "明日方舟"]

# 或者用关键词建议辅助
npx -y @talesofai/neta-skills@latest suggest_keywords --prefix "崩" --size 10
# 输出示例：["崩坏星穹铁道", "崩坏 3", "崩坏学园"]

Step 3: Verify the Category Path

Verify the validity of the path before formal use:

npx -y @talesofai/neta-skills@latest validate_tax_path \
  --tax_path "衍生创作类>同人二创>崩坏星穹铁道"
# 如果有效会返回成功，否则提示错误

Step 4: Get Recommended Content

Use the verified path to get content:

# 精确模式：按分类筛选
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent exact \
  --tax_paths "衍生创作类>同人二创>崩坏星穹铁道" \
  --page_size 20

# 或者搜索模式：结合关键词
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent search \
  --search_keywords "崩坏星穹铁道，同人" \
  --tax_paths "衍生创作类>同人二创" \
  --page_size 20

Common Usage Scenarios

Scenario 1: Exploration with No Clear Goal

The user is just browsing casually with no specific goal.

# 策略：使用推荐模式，广泛浏览
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent recommend \
  --page_size 20

Tips:

• Do not set any filter conditions
• Let the algorithm recommend based on popularity
• Good for discovering unexpected surprises

Scenario 2: Having a Vague Direction of Interest

The user is interested in a certain topic but is unsure about the specific content.

# 步骤 1：先用关键词建议发现方向
npx -y @talesofai/neta-skills@latest suggest_keywords --prefix "游" --size 15

# 步骤 2：基于发现的关键词找标签
npx -y @talesofai/neta-skills@latest suggest_tags --keyword "游戏" --size 15

# 步骤 3：使用搜索模式探索
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent search \
  --search_keywords "原神" \
  --page_size 20

Tips:

• Start with broad keywords
• Gradually narrow the scope
• Combine tags and keywords

Scenario 3: Having a Clear Category Goal

The user clearly wants to see content in a certain category.

# 步骤 1：确认分类路径
npx -y @talesofai/neta-skills@latest suggest_categories --level 1
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "衍生创作类"

# 步骤 2：验证路径
npx -y @talesofai/neta-skills@latest validate_tax_path \
  --tax_path "衍生创作类>同人二创>崩坏星穹铁道"

# 步骤 3：精确筛选
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent exact \
  --tax_paths "衍生创作类>同人二创>崩坏星穹铁道" \
  --page_size 20

Tips:

• Verify the path before using it
• Use exact mode to ensure precise matching
• Multiple category paths can be combined

Scenario 4: Research Before Content Creation

Preparing to create content, need to understand popular tags and categories.

# 步骤 1：了解热门标签
npx -y @talesofai/neta-skills@latest suggest_tags --keyword "角色塑造" --size 20

# 步骤 2：了解相关分类
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "衍生创作类"

# 步骤 3：查看该分类下的热门内容
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent search \
  --search_keywords "角色,设定" \
  --tax_paths "衍生创作类>同人二创" \
  --page_size 30

Tips:

• Gain a comprehensive understanding before creating
• Refer to popular tags to increase exposure
• Choose the appropriate category for placement

Scenario 5: Excluding Specific Content

Wanting to see a certain type of content, but wanting to exclude certain elements.

npx -y @talesofai/neta-skills@latest suggest_content \
  --intent search \
  --search_keywords "AI,绘画" \
  --tax_paths "数字艺术" \
  --exclude_keywords "教程,广告" \
  --exclude_tax_paths "数字艺术>课程培训" \
  --page_size 20

Tips:

• Use exclude_keywords to filter out unwanted content
• Use exclude_tax_paths to exclude specific categories
• Combine positive filtering + negative exclusion

Parameter Combination Tips

Combination 1: Keyword + Category Dual Filtering

npx -y @talesofai/neta-skills@latest suggest_content \
  --intent search \
  --search_keywords "视频,剪辑" \
  --tax_paths "数字艺术>视频制作" \
  --page_size 20

Effect: Search for content containing "视频" and "剪辑" under the "视频制作" category

Combination 2: Multi-level Category Combination

npx -y @talesofai/neta-skills@latest suggest_content \
  --intent exact \
  --tax_paths "衍生创作类>同人二创>崩坏星穹铁道" \
  --page_size 20

Effect: Precisely filter third-level categories, equivalent to tax_paths: "衍生创作类>同人二创>崩坏星穹铁道"

Combination 3: Recommendation Mode + Exclusion Conditions

npx -y @talesofai/neta-skills@latest suggest_content \
  --intent recommend \
  --exclude_keywords "教程,搬运" \
  --exclude_tax_paths "课程类" \
  --page_size 20

Effect: Smart recommendations, but exclude tutorials and reposted content

Combination 4: Pagination Continuity

# 第 1 页
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 0 \
  --page_size 20 \
  --intent search \
  --search_keywords "创意"

# 保存返回的 biz_trace_id（假设返回值为 "abc123"）

# 第 2 页（使用第 1 页返回的 biz_trace_id）
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 1 \
  --page_size 20 \
  --intent search \
  --search_keywords "创意" \
  --biz_trace_id "abc123"

Effect: Maintain consistency of search results

Output Data Structure

suggest_keywords / suggest_tags / suggest_categories

Return a list of suggestions:

{
  "suggestions": ["建议 1", "建议 2", "建议 3"]
}

validate_tax_path

Verify whether the category path is valid:

{
  "valid": true,
  "message": "路径有效"
}

Or when invalid:

{
  "valid": false,
  "message": "错误信息"
}

suggest_content

Return the recommended content list and pagination information:

{
  "module_list": [
    {
      "data_id": "模块 ID",
      "module_id": "模块类型",
      "template_id": "模板 ID",
      "json_data": {}
    }
  ],
  "page_data": {
    "has_next_page": true,
    "page_index": 0,
    "page_size": 20,
    "biz_trace_id": "会话追踪 ID"
  }
}

Performance Optimization Suggestions

1. Set page_size reasonably

• Exploration stage: 10-15 (quick trial and error)
• In-depth browsing: 20-30 (reduce the number of page turns)
• Precise search: 20-40 (get enough content at once)

2. Cache suggestion results

Cache suggestion results such as categories and tags to avoid repeated requests:

# 第一次请求
npx -y @talesofai/neta-skills@latest suggest_categories --level 1 > /tmp/categories_level1.json

# 后续使用缓存（示例：从缓存文件读取）
cat /tmp/categories_level1.json

In actual applications, a file system or database can be used to cache the results.

3. Preload the next-level categories

While the user is viewing the current level, preload possible next-level categories in the background:

# 获取 Level 1 分类
npx -y @talesofai/neta-skills@latest suggest_categories --level 1

# 并行预加载 Level 2 分类（示例：使用后台任务）
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "衍生创作类" &
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "数字艺术" &
npx -y @talesofai/neta-skills@latest suggest_categories --level 2 --parent_path "生活方式" &
wait

This can improve the user experience and reduce waiting time.

4. Batch validate category paths

Validate the validity of multiple category paths at once:

# 并行验证多个路径
npx -y @talesofai/neta-skills@latest validate_tax_path --tax_path "衍生创作类>同人二创>崩坏星穹铁道" &
npx -y @talesofai/neta-skills@latest validate_tax_path --tax_path "衍生创作类>同人二创>原神" &
npx -y @talesofai/neta-skills@latest validate_tax_path --tax_path "数字艺术>概念设计" &
wait

# 或者使用脚本批量处理
cat > /tmp/paths.txt << EOF
衍生创作类>同人二创>崩坏星穹铁道
衍生创作类>同人二创>原神
数字艺术>概念设计
EOF

while read path; do
  npx -y @talesofai/neta-skills@latest validate_tax_path --tax_path "$path"
done < /tmp/paths.txt

This can filter out valid category paths in advance and avoid subsequent request failures.

Debugging Tips

1. Check whether the search keywords are taking effect

# 开启 debug 日志
DEBUG=* npx -y @talesofai/neta-skills@latest suggest_content \
  --intent search \
  --search_keywords "测试关键词"

Check whether the keyword parameters are passed correctly in the logs.

2. Verify category path matching

# 先用 exact 模式测试
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent exact \
  --tax_paths "你的分类路径" \
  --page_size 5

# 如果返回空，可能是：
# - 分类路径不存在
# - 该分类下没有内容
# - 路径格式不正确

3. Compare the results of different modes

# 同一关键词，对比三种模式
npx -y @talesofai/neta-skills@latest suggest_content --intent recommend --page_size 10
npx -y @talesofai/neta-skills@latest suggest_content --intent search --search_keywords "关键词" --page_size 10
npx -y @talesofai/neta-skills@latest suggest_content --intent exact --tax_paths "分类路径" --page_size 10

Observe the differences in results under different modes to understand the characteristics of each mode.

Common Questions

Q1: What is the difference between suggest_keywords and suggest_tags?

A:

• suggest_keywords: Based onprefixfuzzy matching, suitable for the exploration stage
• suggest_tags: Based oncomplete keywordsrelevance matching, more precise

# keywords - 前缀匹配
npx -y @talesofai/neta-skills@latest suggest_keywords --prefix "崩"  # 返回所有以"崩"开头的词

# tags - 相关性匹配
npx -y @talesofai/neta-skills@latest suggest_tags --keyword "游戏"   # 返回与"游戏"相关的标签

Q2: Why does validate_tax_path pass validation, but suggest_content returns empty?

Possible reasons:

1. The category path is valid, but there is temporarily no content under that category
2. The intent mode is incorrect (exact should have been used but recommend was used)
3. There are other conflicting filter conditions

Solution:

# 尝试去掉其他条件，只用分类路径
npx -y @talesofai/neta-skills@latest suggest_content \
  --intent exact \
  --tax_paths "你的分类路径" \
  --page_size 10

Q3: How do I choose the appropriate intent mode?

Selection guide:

Mode	Use case	Required parameters
recommend	Aimless browsing, exploration	None
search	Has clear keywords	search_keywords
exact	Strict filtering by category	tax_paths or category parameters

Q4: How do I use exclude_keywords and exclude_tax_paths?

# 排除包含特定词的内容
npx -y @talesofai/neta-skills@latest suggest_content \
  --exclude_keywords "教程,广告,搬运" \
  --search_keywords "绘画"

# 排除特定分类
npx -y @talesofai/neta-skills@latest suggest_content \
  --exclude_tax_paths "课程培训,商业推广" \
  --tax_paths "数字艺术"

Note: Exclusion conditions will significantly reduce the number of results, so use them with caution.

Q5: How is biz_trace_id maintained across multiple rounds of dialogue?

Method: Save the biz_trace_id returned by the first request, and use this value for all subsequent requests.

# 第 1 次请求（首页）
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 0 \
  --page_size 20 \
  --intent search \
  --search_keywords "关键词" > /tmp/page0.json

# 从返回结果中提取 biz_trace_id
BIZ_TRACE_ID=$(cat /tmp/page0.json | jq -r '.page_data.biz_trace_id')

# 第 2 次请求（下一页），使用保存的 biz_trace_id
npx -y @talesofai/neta-skills@latest suggest_content \
  --page_index 1 \
  --page_size 20 \
  --intent search \
  --search_keywords "关键词" \
  --biz_trace_id "$BIZ_TRACE_ID"

Note: You should use the biz_trace_id returned by the first request, rather than using the return value from the previous request each time.

Summary

Key points for using the gameplay content exploration skill:

1. Progressive exploration: category → tag → validation → content
2. Mode selection: use recommend for aimless browsing, search when there are keywords, and exact when there is a category
3. Parameter combination: make good use of dual filtering with keywords + categories
4. Exclusion techniques: use exclude parameters reasonably to filter out noise
5. Conversation continuity: Properly maintain biz_trace_id to ensure a consistent experience

By following these best practices, you can efficiently explore and discover high-quality content on the platform!