drivable-vscode

Operate a running VS Code instance remotely for exploratory testing and evidence capture.

1.0k|454|Updated Jun 21, 2017
One-click install
npx skills add https://github.com/forcedotcom/salesforcedx-vscode --skill drivable-vscode
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: drivable-vscode
Source: https://github.com/forcedotcom/salesforcedx-vscode/tree/main/.claude/skills/drivable-vscode
Command: npx skills add https://github.com/forcedotcom/salesforcedx-vscode --skill drivable-vscode

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Manually reproducing customer bugs and verifying VS Code extension behavior requires repetitive clicking, screenshotting, and note-taking. This Skill lets an agent drive a real VS Code instance through an MCP server, observing the UI and acting on it step by step while recording findings and evidence.

Core Features & Use Cases

  • Remote UI Operation: Start a VS Code session, observe screenshots plus ARIA/text state, and perform actions through MCP tools like start, observe, act, and waitForText.
  • Findings and Evidence Capture: Record defects with add_finding and finish sessions with saved screenshots, video, action logs, console logs, and summaries.
  • Scripted and Agent-Driven Workflows: Run repeatable scenario matrices via a validated example script, or pursue open-ended exploratory testing and bug reproduction.
  • Use Case: Reproduce a reported deploy-command failure from a clean VS Code session, document each defect with reproduction steps, and finish with screenshot and video evidence.

Quick Start

Use drivable-vscode to reproduce the reported bug in a clean VS Code session, record each defect as a finding, and finish with screenshot and video evidence.

Frequently Asked Questions about drivable-vscode

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate VS Code UI testing with an AI agent?

Register the drivable-vscode MCP server with your client, then call start with a goal, observe the screenshot and ARIA state, act on one element at a time, and wait for expected text transitions. Finish the session to save screenshots, video, logs, and findings.

How do I set up the drivable-vscode MCP server in Claude Code?

Run npm install and compile the drivable-vscode workspace, then add the server with claude mcp add pointing to the drivable-vscode-mcp.js binary. Restart Claude Code and check /mcp if tools are unavailable.

Can drivable-vscode run scripted repeatable test scenarios?

Yes, a scripted workflow supports repeatable checks and scenario matrices by driving MCP with a client and asserting observations. A validated example script is provided at packages/drivable-vscode/scripts/drivable-vscode-example.mjs.

Does drivable-vscode work with Cursor and OpenCode?

Yes, Cursor accepts the server via cursor --add-mcp or .cursor/mcp.json, and OpenCode via an mcp entry in opencode.jsonc. Both require a restart before the tools become available.

What are the limitations of drivable-vscode?

One server process owns at most one active session, and native OS dialogs and external windows are outside its control. Screenshots and video do not redact credentials, so avoid displaying secrets during sessions.