Skip to main content
Skills are Markdown files that provide your agents with reusable workflows and context for when to use their tools. Agent Studio maintains a catalog of skills that any agent can use. Add a skill once, then enable it for individual agents from their configuration page.

​
Viewing available skills

Select Skills from the left-hand navigation to view your skill catalog. Skills appear as cards showing the skill name and a short description.
Skills page showing a grid of skill cards, each displaying a skill name and short description
Click a skill’s card to open its details dialog. The dialog has two tabs:
  • About: Displays the full skill content.
  • Security: Runs a security scan on the skill’s files.
From the details dialog you can also click Delete to remove the skill, or Add to Agents to enable it for one or more agents.

​
Adding a skill

  1. Select Skills from the left-hand navigation.
  2. Click Add in the upper-right corner.
  3. Choose a source and configure the skill:
    Add a skill from a public GitHub repository.
    1. Enter the repository Owner and Repository name.
    2. (Optional) Enter a Skill name to add a specific skill from the repository. Leave this field blank to add all skills in the repository.
    3. Click Add.
    Add Skill dialog showing the GitHub tab with Owner, Repository, and Skill name fields

​
Scanning skills for security threats

Agent Studio can scan skills for security vulnerabilities using static analysis and optional AI-assisted reasoning. Scans identify potential threats such as dangerous code execution patterns, suspicious behavior, and injection risks.
Scan results are best-effort and are not a certification of safety.
To scan a skill:
  1. Select Skills from the left-hand navigation.
  2. Click the skill’s card to open its details dialog.
  3. Select the Security tab.
  4. (Optional) Enable Deep scan with AI for an additional semantic pass that sends the skill’s content to your configured AI provider.
  5. Click Start scan.
To run the scan again, click Re-scan.

​
Severity levels

The scan summary displays the highest severity found across all findings:

Critical

Immediate security risk requiring attention.

High

Significant risk that should be addressed before use.

Medium

Potential risk worth reviewing (for example, use of subprocess.run with untrusted input).

Low

Minor concern with limited risk.

None

No issues found.

​
Deep scan with AI

By default, scans use offline analyzers (static analysis, bytecode inspection, and behavioral analysis). Enabling Deep scan with AI adds a semantic pass that uses your configured AI provider to reason about the skill’s intent.
Deep scan with AI requires a configured AI provider with a valid API key. If no compatible provider is available, the toggle is disabled.

​
Enabling a skill for an agent

Once a skill is in the catalog, you can enable it for individual agents from the agent’s configuration page. See Skills for instructions.