CORE CONCEPTS5 min read
Skills
Creating reusable prompt templates for AI assistants
What are Skills?
Skills are markdown documents with YAML frontmatter that get exposed as MCP resources. They let you define custom behaviors, prompts, or templates that AI assistants can use when working with your data.
Skill Format
markdown
--- title: Explain Sales Data argument-hint: The product or category to explain --- # Sales Explanation You are analyzing sales data for: **$ARGUMENTS** ## Instructions 1. Look at the sales trends for the specified product 2. Identify key metrics: revenue, units sold, growth rate 3. Explain any anomalies or interesting patterns 4. Provide actionable recommendations ## Context The data comes from our ERP system and is refreshed daily. When explaining numbers, always compare to the previous quarter.
INFOThe $ARGUMENTS placeholder is replaced with the user's input when the skill is invoked.
Frontmatter Fields
| Field | Required | Description |
|---|---|---|
| title | Yes | Display name for the skill |
| argument-hint | No | Hint for the input parameter |
Exposing Skills
Skills are exposed in two ways through MCP servers:
As a Tool
Skills appear as callable tools with an "arguments" parameter. When called, $ARGUMENTS is replaced and the markdown is returned.
As a Resource
Skills are available at skill://skill-name for reading directly.
AI Generation
Conexor can generate skills automatically from a description:
- Go to Dashboard → Skills
- Click "Generate with AI"
- Describe what the skill should do
- Review and edit the generated markdown
- Save and assign to MCP servers
Best Practices
- Keep skills focused on a single task
- Use clear, specific instructions
- Include context about your data schema
- Test skills with Claude before deploying
- Version control your skills alongside your code
TIPSkills work great for standardizing how your team interacts with data through AI.