The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/setup script modifies the user's shell configuration files (e.g., ~/.bashrc, ~/.zshrc, ~/.profile) to persistently add ~/.local/bin to the system PATH. It also creates symlinks in this directory.
[COMMAND_EXECUTION]: The skill uses browser-harness-js to execute dynamic, runtime-generated JavaScript code within a browser context. This allows the skill to programmatically control browser behavior, such as navigation and scrolling, and to evaluate DOM queries via the Chrome DevTools Protocol (CDP).
[DATA_EXFILTRATION]: The skill is designed to interact with and extract data from an active, authenticated X (Twitter) session in the user's browser. While this is the primary function of the skill, it leverages the user's credentials and session state to bypass login walls.
[PROMPT_INJECTION]: The skill ingests and processes search results and tweet text from X.com, which is untrusted user-generated content. This creates an indirect prompt injection surface where malicious content encountered during a search could attempt to manipulate the agent's instructions or output.
Ingestion points: Data enters the agent's context through DOM extraction of tweets from X.com search results and permalink pages.
Boundary markers: None identified. Extracted content is presented to the agent as raw text.
Capability inventory: The skill can execute shell scripts and manipulate a local browser via CDP.
Sanitization: The extracted text is not sanitized or filtered before being returned to the agent.

xsearch