# Skills (/docs/verdent-manager/core-features/skills)
> Extend Verdent's capabilities with reusable knowledge packs for specialized workflows and domain expertise
Verdent **Skills** are reusable knowledge packs that extend the Agent's expertise. Each Skill contains domain-specific prompts, scripts, references, and best practices. Simply invoke a Skill by name, and the Agent loads the corresponding context and follows the predefined workflow.
***
## What Are Skills? [#what-are-skills]
### Core Concept [#core-concept]
A **Skill** is a structured knowledge pack stored in a folder. Every Skill contains a `SKILL.md` file that defines its name, description, and detailed prompt. When you invoke a Skill in conversation, Verdent reads this file and injects its content into the Agent's context.
### How Skills Differ from Subagents and MCP [#how-skills-differ-from-subagents-and-mcp]
| Dimension | Skills | Subagents | MCP |
| ------------------ | -------------------------------------------- | ----------------------------------------------------- | -------------------------------------------------- |
| **Trigger** | Manual invocation via `@skill` or mention | Auto-spawned by the main Agent | Auto-invoked or explicitly called |
| **Execution** | Expands the main conversation context | Runs as an independent sub-task with isolated context | Calls external tools via JSON-RPC 2.0 |
| **Purpose** | Domain guidance (e.g., "how to write tests") | Independent operations (e.g., "run code review") | External data/tool access (e.g., database queries) |
| **Lifecycle** | One-time context injection | Completes sub-task then returns results | Persistent server process during session |
| **Implementation** | Markdown files + optional scripts | Built-in Agent logic | Executable server processes |
In short: **Skills provide knowledge and workflow guidance**, **Subagents execute independent tasks**, and **MCP connects external tools and data sources**.
***
## How to Use Skills [#how-to-use-skills]
### Invoking a Skill [#invoking-a-skill]
You can invoke a Skill in two ways:
**1. Mention the Skill name in conversation**
```
Use the skill-creator skill to help me create a new Skill
```
**2. Use the @ mention syntax**
```
@skill-creator help me create a Skill for API documentation generation
```
When a Skill is triggered:
1. The Agent calls the `skill` tool to read the Skill's `SKILL.md` file
2. The file content is injected into the current conversation context
3. The Agent continues working according to the Skill's guidance
### Browsing Available Skills [#browsing-available-skills]
Access the Skills panel through Settings:
1. Open the **Settings** menu
2. Select the **Skills** tab
3. Browse installed Skills and available Skills in the Skill Store
You can also type `@` in the input box to trigger autocomplete and quickly select an available Skill.
***
## Built-in Skills [#built-in-skills]
Verdent ships with a set of built-in Skills covering common workflows:
| Skill | Description |
| ------------------- | ---------------------------------------------------------------------------------------------- |
| **docx** | Read, create, or edit Word documents (`.docx`) with format-faithful visual checks |
| **find-skills** | Discover and install Skills — use when asking "how do I do X?" or looking for new capabilities |
| **frontend-design** | Create distinctive, production-grade frontend interfaces with high design quality |
| **pdf** | Read, create, or review PDF files with visual rendering and content extraction |
| **pptx** | Read, create, or edit PowerPoint presentations (`.pptx`) |
| **skill-creator** | A guided workflow for creating and iterating on custom Skills |
| **xlsx** | Read, analyze, visualize, and smart-edit Excel spreadsheets |
More Skills are available in the **Skill Store**. Open **Settings → Skills → Store** to browse and install additional Skills.
***
## Creating Custom Skills [#creating-custom-skills]
You can create custom Skills to encapsulate your team's domain expertise or project-specific workflows.
### Skill Directory Structure [#skill-directory-structure]
A standard Skill folder looks like this:
```
my-custom-skill/
├── SKILL.md # Required: Skill definition file
├── agents/
│ └── verdent.yaml # Optional: UI configuration (icons, display name)
├── scripts/ # Optional: helper scripts
└── references/ # Optional: reference docs or examples
```
### SKILL.md Format [#skillmd-format]
`SKILL.md` is the core Skill file, containing YAML frontmatter and a Markdown body:
```markdown
---
name: my-custom-skill
description: A concise description shown in the Skill list
metadata:
version: "1.0.0"
author: "Your Name"
license: "MIT"
---
# Detailed Instructions
This is the detailed prompt content the Agent reads.
## Workflow
1. Step one
2. Step two
3. ...
## Best Practices
- Practice suggestion 1
- Practice suggestion 2
```
**Frontmatter field requirements:**
* `name` (required): Must match the folder name. Only lowercase letters, numbers, and hyphens (`a-z0-9-`) allowed. Length: 1–64 characters. No consecutive hyphens or leading/trailing hyphens.
* `description` (required): A short description, max 1024 characters.
* `metadata` (optional): Version, author, license, and other meta information.
### Installing Custom Skills [#installing-custom-skills]
**Method 1: Import via Settings**
1. Open **Settings → Skills**
2. Click **Import Skill**
3. Select the Skill folder (or a `.zip` / `.skill` archive)
4. Verdent validates `SKILL.md` and installs it to `~/.verdent/skills/`
**Method 2: Manual copy**
```bash
cp -r my-custom-skill ~/.verdent/skills/
```
The Skill becomes available after restarting Verdent or refreshing the Skills list.
**Method 3: Project-level Skill**
Place the Skill inside your project directory:
```bash
cp -r my-custom-skill /path/to/your/project/.verdent/skills/
```
Project-level Skills are only visible within that project and take priority over global Skills with the same name.
***
## Skill Scopes [#skill-scopes]
Verdent supports three levels of Skill scoping:
| Level | Location | Visibility |
| ----------- | ---------------------------- | --------------------------------- |
| **Global** | `~/.verdent/skills/` | All projects and workspaces |
| **Project** | `/.verdent/skills/` | Current project's workspaces only |
**Priority rules:**
When Skills with the same name exist at multiple scopes:
* Project overrides Global
***
## Skill Store [#skill-store]
Verdent provides a **Skill Store** for browsing and installing community and official Skills.
### Browsing the Skill Store [#browsing-the-skill-store]
1. Open **Settings → Skills**
2. Switch to the **Store** tab
3. Browse available Skills or use the search box
### Installing from the Store [#installing-from-the-store]
1. Find the desired Skill in the Store
2. Click **Install**
3. The Skill is automatically downloaded and installed to `~/.verdent/skills/`
**Security verification:**
Skills installed from the Store undergo SHA256 checksum verification to ensure file integrity and security.
***
## Practical Examples [#practical-examples]
### Example 1: Creating a New Skill with `skill-creator` [#example-1-creating-a-new-skill-with-skill-creator]
```
@skill-creator help me create a Skill for guiding the team on writing Go unit tests
```
The Agent will:
1. Read the `skill-creator` Skill content
2. Guide you through filling in the Skill name, description, and core prompt
3. Generate a standard `SKILL.md`
4. Suggest installation path and verification steps
### Example 2: Analyzing Data with `spreadsheet` [#example-2-analyzing-data-with-spreadsheet]
```
@spreadsheet read sales-2025.xlsx from the project root, analyze Q1 sales trends, and generate a chart
```
The Agent will:
1. Load the `spreadsheet` Skill
2. Use `pandas` and `openpyxl` to read the Excel file
3. Analyze the data and generate visualizations
4. Save results or display them in the conversation
### Example 3: Fixing CI with `gh-fix-ci` [#example-3-fixing-ci-with-gh-fix-ci]
```
@gh-fix-ci my PR #123 GitHub Actions tests are failing, help me debug
```
The Agent will:
1. Use `gh` CLI to retrieve CI logs for PR #123
2. Analyze the failure (e.g., test case errors, dependency issues)
3. Propose a fix
4. With your approval, modify the code and push the fix
***
## FAQs [#faqs]
When a Skill is invoked, the Agent reads the `SKILL.md` file, which counts toward token usage and therefore consumes credits. However, the Skill itself does not make additional API calls.
Yes. Install the Skill to `~/.verdent/skills/` (global scope) and it will be available in all projects.
In **Settings → Skills**, right-click the Skill and select **Delete**. Built-in Skills marked as `undeletable` cannot be deleted and will be reinstalled on next startup.
Yes. Verdent automatically detects Skills in `~/.cursor/skills`, `~/.claude/skills`, and `~/.codex/skills`, and prompts you to import them. Click **Import from External Sources** in **Settings → Skills**.
* **Project Rule**: Always-active instructions automatically injected into every conversation
* **Skill**: On-demand knowledge packs injected only when invoked
Use Rules for universal coding standards; use Skills for specific domain workflows.
Skills installed from the Skill Store are automatically checked for updates in the background. Manually installed Skills need to be re-imported (overwriting the old version) to update.
***
## Advanced Usage [#advanced-usage]
### Embedding Scripts in Skills [#embedding-scripts-in-skills]
Add a `scripts/` directory to your Skill and reference the scripts in `SKILL.md`:
```markdown
## Data Processing Script
Run the following command to process data:
\`\`\`bash
python scripts/process_data.py --input data.csv --output results.json
\`\`\`
```
The Agent reads this instruction and executes the script as needed.
### Referencing Documentation [#referencing-documentation]
Store API docs, specs, or sample code in a `references/` directory and link them in `SKILL.md`:
```markdown
## Reference Documentation
See [references/api-spec.md](references/api-spec.md) for the detailed API specification.
```
The Agent can read these reference files based on the Skill's instructions.
***
## Best Practices [#best-practices]
**Keep Skills focused.** Each Skill should target a single domain or workflow. Avoid creating "do-everything" Skills—split them into smaller, specialized ones instead.
**Use clear naming.** Skill names should be concise and descriptive, e.g., `api-doc-generator` rather than `my-skill-1`.
**Provide examples and references.** Include concrete examples and reference links in `SKILL.md` to help the Agent better understand the expected output.
**Maintain version numbers.** Use the `metadata.version` field to track updates and compatibility.
**Test before sharing.** Validate your Skill in a test project before rolling it out to the team or publishing to the Skill Store.
***
## See Also [#see-also]
How subagents work and how to manage them
Connect external tools and services via MCP
Configure always-active Project and User Rules
Review code changes with the built-in Reviewer