Skills¶

Skills are reusable working methods or professional capabilities for Agents. After a skill is enabled, the Agent can handle tasks according to predefined procedures, standards, and output formats, such as alert triage, log investigation, trace analysis, inspection report generation, change risk checks, or incident review organization.

Skills follow a "prepare globally first, then enable per Agent" configuration model. Prepare the skills available to the team in Global Configuration, then return to a specific Agent workspace and enable only the skills that Agent needs. This avoids giving every Agent the same capabilities by default and helps keep each Agent's responsibility boundary clear.

Configuration Flow¶

The recommended configuration sequence is:

flowchart TD
  A[Open Global Configuration > Skills] --> B{Choose a skill source}
  B --> C[Install from Skill Marketplace]
  B --> D[Create a custom skill]
  B --> E[Upload a ZIP package with SKILL.md]
  C --> F[Build the workspace skill pool]
  D --> F
  E --> F
  F --> G[Open the Agent workspace > Skills]
  G --> H[Enable marketplace or custom skills by Agent responsibility]
  G --> I[Built-in skills are available by default]
  H --> J[Use skills in tasks]
  I --> J

Install Skills From The Marketplace¶

The Skill Marketplace provides skills that can be installed directly. In Global Configuration > Skills, browse available skills in Marketplace or use search to find a specific one.

When viewing marketplace skills, you can:

• Click a skill card to view details, including its description, tags, and version.
• Click Install to add the skill to the current workspace skill pool.
• After installation, view the skill under Installed and filter by built-in, marketplace, or custom source.

Installing a marketplace skill does not automatically enable it for every Agent. It only adds the skill to the workspace skill pool. You still need to enable it in the specific Agent workspace.

Create Or Import Custom Skills¶

If your team needs to standardize its own working methods, create custom skills in Global Configuration. When adding a skill, you can choose:

• Create blank skill: enter the skill name, description, icon, and skill documentation.
• Upload ZIP: upload a skill package that contains SKILL.md; the system parses it and creates the skill.

For a blank skill, it is recommended to define:

• Skill name: use an easy-to-recognize name, such as code-review-helper or test-case-generator.
• Description: explain what problem the skill helps solve, so the team can decide whether to enable it.
• Skill documentation: use Markdown to describe applicable scenarios, input requirements, processing steps, output standards, and examples.

In the installed skill list, click a skill card to view its details. For uploaded skills, the detail view shows the package contents, helping you confirm that SKILL.md and related files are correct.

Enable Skills For An Agent¶

Open a specific Agent workspace and go to Skills. This page shows the skills already prepared in the workspace, grouped by Built-in, Marketplace, and Custom.

Keep the following in mind:

• Built-in skills are available to the Agent by default and cannot be manually disabled.
• Marketplace skills and custom skills are enabled with switches. Changes are saved immediately for the current Agent.
• This page only controls Agent-level enablement. To add, edit, or delete skills, click Global Configuration and manage them there.
• Click the body of a skill card to open a read-only detail view and inspect the skill's instructions or package contents.

Enable skills according to the Agent's responsibility. For example, an alert on-call Agent can enable alert triage and notification summary skills, a production troubleshooting Agent can enable log, metric, and trace analysis skills, and an inspection Agent should only enable inspection report and risk summary skills.

Use Skills In Tasks¶

After a skill is enabled, the Agent can choose a suitable skill based on the task goal. You can also explicitly specify which skill you want the Agent to use in the task description.

When a skill affects the Agent's process or output format, describe the goal, input scope, and expected result clearly. For example:

• Ask the Agent to output impact scope, possible causes, and recommended actions according to an alert handling standard.
• Ask the Agent to summarize logs, metrics, and trace clues related to an incident and suggest an investigation order.
• Ask the Agent to generate inspection reports, incident reviews, or change risk notes according to the team's template.

Best Practices¶

• Prepare skills as shared team capabilities in Global Configuration, then enable them per Agent responsibility.
• Enable only responsibility-relevant skills for each Agent to keep behavior stable and predictable.
• Write the skill description for users, so team members can quickly understand when to use it.
• Write the skill documentation for Agent execution, focusing on inputs, steps, constraints, output format, and examples.
• After importing a skill package, click the card to inspect its details before enabling it in an Agent workspace.