Skip to content

[FEAT] Support for Workspace Environments and Agent Skills #1008

@trfi

Description

@trfi

Is your feature request related to a problem? Please describe.

I would like to propose the implementation of a Workspace system and support for Agent Skills to enhance VoltAgent's ability to handle persistent files, execute code securely, and reuse complex task instructions.

Currently, agents often operate in a stateless or memory-limited context. Introducing a Workspace will provide agents with a persistent environment to store files, execute shell commands, and search through indexed content.

Describe alternatives you've considered

No response

Additional context

By adopting the Agent Skills open standard, VoltAgent stays compatible with the evolving agentic workflow ecosystem.

Describe the thing to improve

Proposed Features

1. Workspace Environment

The Workspace should act as the agent's "home base" during a task, providing:

  • Filesystem: Standard operations (Read, Write, List, Delete).
  • Search: Indexed search capabilities (BM25, Vector, or Hybrid) over the workspace files.
  • Sandbox Integration: A controlled environment for code execution.

2. Sandbox Providers

To ensure security and flexibility, the workspace should support multiple sandbox providers for command execution. Key requirements include command timeouts, environment variable control, and OS-level isolation.
Proposed Providers to support:

  • Local (Default)
  • E2B / Daytona / Modal
  • Cloudflare Sandboxes / Vercel Sandbox

3. Agent Skills Support

Integration with the Agent Skills specification.

  • Skills are reusable, versioned instructions that teach agents how to perform specific tasks.
  • Supporting this open standard will allow VoltAgent to import and export capabilities seamlessly across the ecosystem.

Use Case

  1. Data Analysis: An agent writes a CSV to the workspace, installs a python library via the Sandbox, and generates a chart.
  2. Coding Assistant: An agent clones a repository, searches the codebase, and runs tests.
  3. Extensibility: Users can drop a new skill folder into the system. The agent will detect the new SKILL.md metadata and load the full instructions only when needed.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions