home / skills / xiaomi / mone / knowledge

This skill helps you locate WeKnora knowledge base guides and best practices with targeted tag filtering and hybrid search.

npx playbooks add skill xiaomi/mone --skill knowledge

Review the files below or copy the command above to add this skill to your agents.

Files (3)
SKILL.md
2.4 KB
---
name: knowledge
description: Search WeKnora knowledge base for documentation, guides, and best practices. Use when user asks "how to", "what is", or needs configuration guides, feature docs, usage instructions, or troubleshooting steps.
---

# Knowledge Base Skill

Search the WeKnora knowledge base with intelligent tag filtering for precise results. Supports hybrid search (keyword + vector).

## When to Use

- User asks "how to use", "how to configure", "what is"
- Questions about product features, usage guides, API docs
- Best practices, configuration examples
- Keywords: usage, configuration, feature, guide, document, how to, 如何, 怎么, 配置, 使用

## Workflow

1. **Get tags first** - Understand available knowledge categories
2. **Match tags** - Select appropriate tags based on user query
3. **Search** - Use tag filtering for better precision

## Scripts

### 1. Get Available Tags

Fetch all tags to understand knowledge structure:

```bash
uv run python .claude/skills/knowledge/scripts/get_tags.py
```

**Output**: JSON with `tags` array containing `name` and `description` for each tag

### 2. Search Knowledge

Search with optional tag filtering:

```bash
uv run python .claude/skills/knowledge/scripts/search.py -q "your query" --tags '["tesla"]'
```

| Parameter | Required | Description |
|-----------|----------|-------------|
| `-q, --query` | Yes | Search query |
| `-t, --tags` | No | JSON array of tag names, e.g., `'["tesla", "miline"]'` |
| `-m, --tag-mode` | No | `any` (default) or `all` |
| `--kb-id` | No | Knowledge base ID (has default) |

**Output**: JSON with `results` array containing `content`, `score`, `metadata`

## Tag Mapping Guide

| User Query Keywords | Suggested Tags |
|---------------------|----------------|
| Tesla, gateway, routing, 网关 | `tesla` |
| Dayu, microservice, governance, 微服务 | `Dayu` |
| CI/CD, pipeline, deployment, 部署, 流水线 | `miline` |
| Task scheduling, cron, 调度 | `moon` |
| General questions, 常见问题 | `常见问题` |

## Examples

### Search with tag filtering (recommended)
```bash
uv run python .claude/skills/knowledge/scripts/search.py -q "如何配置网关" --tags '["tesla"]'
```

### Search with multiple tags
```bash
uv run python .claude/skills/knowledge/scripts/search.py -q "部署流水线" --tags '["miline"]' --tag-mode any
```

### Search without tags (fallback)
```bash
uv run python .claude/skills/knowledge/scripts/search.py -q "系统架构设计"
```

Overview

This skill searches the WeKnora knowledge base to find documentation, configuration guides, feature descriptions, and troubleshooting steps. It combines keyword and vector search with intelligent tag filtering to return precise, relevant results. Use it when you need actionable how-to instructions or authoritative product docs. The skill exposes simple scripts to list tags and run filtered searches.

How this skill works

First it fetches available knowledge tags to understand the knowledge base structure. It maps user intent to one or more tags, then runs a hybrid search (keyword + vector) with optional tag filters to improve precision. Results are returned as JSON containing content, relevance score, and metadata for easy consumption or display.

When to use it

  • You ask “how to” or “how do I” about product setup, configuration, or usage
  • You need feature documentation, API references, or usage examples
  • You want troubleshooting steps, common issues, or best practices
  • You need targeted results for a specific area by filtering with tags

Best practices

  • Get the list of available tags first to guide precise queries
  • Map user keywords to suggested tags (e.g., tesla, miline, moon) before searching
  • Use tag-mode 'any' for broader results or 'all' for strict intersection
  • Include clear, compact queries to leverage vector search effectively
  • Fallback to an untagged search when tag mapping is uncertain

Example use cases

  • Find a gateway configuration guide by mapping keywords to the 'tesla' tag
  • Search CI/CD pipeline examples using the 'miline' tag and deployment keywords
  • Locate task scheduling documentation by filtering with the 'moon' tag
  • Retrieve common troubleshooting steps by querying the '常见问题' tag
  • Run an untagged search for general architecture or design questions

FAQ

How do I list all available tags?

Run the provided tags script to fetch a JSON array of tag names and descriptions before searching.

Can I search with multiple tags?

Yes — pass a JSON array of tag names and choose tag-mode 'any' for union or 'all' for intersection.

What if tag mapping is unsure?

Use a broader query without tags as a fallback; the hybrid search will still return relevant results.