Skills Management

Skills are the Agent's "capability modules." A skill defines the Agent's behavior in specific scenarios, the tools used, and execution steps. By adding skills, you can quickly extend the Agent's capabilities.

What are Skills

Skills are essentially structured instruction documents (SKILL.md) that tell the Agent what to do when facing a certain type of task.

For example, a "Contract Review" skill might include:

• Trigger Condition: When user submits a contract document
• Execution Steps: Check clauses one by one, identify risk points, provide modification suggestions
• Tools Used: Document parsing, legal knowledge base retrieval
• Output Format: Structured review report

Difference between Skills and Tools:

Dimension	Skill	Tool
Essence	High-level behavior instruction	Low-level capability interface
Granularity	A complete task workflow	A specific operation
Example	"Contract Review"	"PDF Parsing", "Text Search"
Contains	Can reference multiple tools	Independent capability unit

Viewing Agent Skills

1. Enter the Agent details page
2. Click the "Skills" tab
3. View the list of installed skills

Each skill card displays:

• Skill name and icon
• Brief description
• Risk level (Low / Medium / High)
• Version information

Click the skill card to view detailed skill documentation (SKILL.md).

Adding and Removing Skills

Installing from Skill Marketplace

1. Enter the Agent's skills management page
2. Click "Add Skill"
3. Browse or search the skill marketplace
4. Click "Install" to add the skill to the current Agent

Removing Skills

1. Find the skill to remove in the skills list
2. Click the menu button on the right side of the skill
3. Select "Remove"
4. After confirmation, the skill file is deleted

Removal is Recoverable

Since Agents use Git to manage files, you can still recover through version history after removing a skill.

Skill Risk Levels

Each skill is marked with a risk level to help you understand the types of operations it involves:

Risk Level	Meaning	Example
Low Risk	Only reads data, does not modify system state	Information retrieval, text analysis
Medium Risk	Requires network access or external service connection	API calls, online search
High Risk	Involves sensitive data operations or system modifications	File writing, command execution

High-risk skills require your explicit confirmation each time they are executed.

Built-in Skills vs. Custom Skills

Built-in Skills

DesireCore comes with a set of Global Skills that all Agents can use. These skills are updated with the client version and include basic capabilities such as:

• Document reading and writing
• Web search
• Code execution

Custom Skills

Advanced users can manually create skills. Each skill is a directory containing:

<skill_id>/
├── SKILL.md          # Required: Skill description + metadata
├── skill.yaml        # Optional: Dependency declaration
├── scripts/          # Optional: Executable scripts
├── references/       # Optional: Reference documents
├── assets/           # Optional: Templates and resources
└── evals/            # Optional: Evaluation test cases

SKILL.md is Core

All skill behaviors are defined by instructions in SKILL.md. Writing a good SKILL.md is equivalent to teaching the Agent a new capability.

Next Steps

• Version Control — Learn about Git version control for Agents
• Agent Marketplace — Discover more skills from the marketplace