X

Domain: Workflow

The X environment simulates a short-form social microblogging platform built around posts, replies, reposts, likes, follows, and personalized feeds. It supports account login, timeline browsing, profile inspection, trend and topic exploration, and post-level interaction, making it suitable for evaluating agents in information-consumption and social-interaction scenarios. This environment is particularly important because public social posts mix first-party user content with third-party quotes, links, hashtags, and replies, creating realistic opportunities for both benign engagement and adversarial manipulation through injected instructions, misleading claims, or socially coercive phrasing.

GUI. Representative GUI views of the simulated X environment are shown in the figure, covering login, profile inspection, topic exploration, and home-timeline browsing.

MCP Tools. The X environment exposes a comprehensive MCP interface for microblogging and social-graph workflows. As summarized in the MCP-tool table, the tool set covers authentication and profile management, social-graph operations, tweet authoring and retrieval, engagement actions (likes, retweets, bookmarks), direct messaging, content discovery (search, trends, mentions), and safety controls (blocking and muting). These tools allow agents to browse timelines, publish and delete tweets, react to content, search for users and hashtags, exchange private messages, and triage mentions in a realistic social setting. Because X conversations intermix user-generated text, media references, external URLs, and cross-user mentions, the X environment is especially useful for evaluating whether agents can safely operate over social-media artifacts without being manipulated by malicious content embedded in tweets, DMs, or trending discussions.

Screenshots

Login page

Profile page

Explore page

Home page

Simulated X environment. Representative views used for account access, profile inspection, topic exploration, and timeline browsing in workflow-agent evaluation.