Prompt Switchboard Compare Workflows

Teach an agent to install Prompt Switchboard's local MCP sidecar, connect it in a host, and run a compare-first browser workflow.

This package is internally coherent for wiring a local Prompt Switchboard MCP sidecar, but it instructs you to clone and run a third‑party GitHub repo and to run npm install / npm run mcp:server locally. Before installing or running anything: 1) inspect the repository and package.json (especially the 'mcp:server' script) to confirm there are no unexpected postinstall scripts or network exfiltration code; 2) prefer a pinned commit or an official/release tarball from a trusted upstream where possible; 3) run npm install and the server in a sandbox/container or on a non‑privileged account, not as root; 4) review the browser extension being used (Prompt Switchboard) and only enable it if you trust its source; 5) if you need stronger assurance, ask the author for signed releases or a maintainer statement. If the repo proves to be an official, audited upstream or a verified package, this concern would drop and the skill would be much safer.

Type: OpenClaw Skill Name: prompt-switchboard-compare-workflows Version: 1.1.1 The skill bundle provides a legitimate integration for "Prompt Switchboard," a tool designed to compare AI model outputs across browser tabs using the Model Context Protocol (MCP). The package includes clear instructions in SKILL.md and references/INSTALL.md for cloning a specific GitHub repository (multi-ai-sidepanel) and running a local MCP server via npm. The workflow is well-defined, focusing on checking bridge status, verifying tab readiness, and performing prompt comparisons. No evidence of data exfiltration, malicious code execution, or harmful prompt injection was found; the bundle's logic and instructions are entirely consistent with its stated purpose of enhancing browser-based AI workflows.