The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/setup script performs persistent system modifications by appending PATH exports to user shell configuration files (~/.zshrc, ~/.bashrc, and ~/.profile). This is used to ensure the gsearch and browser-harness-js binaries are available in future shell sessions.
[COMMAND_EXECUTION]: The scripts/gsearch script employs dynamic code generation by using node -e to perform string replacements on a JavaScript template. The resulting script is then piped into the browser-harness-js utility for execution.
[PROMPT_INJECTION]: The skill represents an indirect prompt injection surface as it retrieves and processes arbitrary content from the internet through Google searches and the follow command.
Ingestion points: External data enters the agent context via the Runtime.evaluate call in scripts/gsearch, which extracts innerText from search results and visited web pages.
Boundary markers: The skill does not provide clear delimiters or instructions to the agent to treat the retrieved web content as untrusted data.
Capability inventory: The skill can execute shell commands, modify local configuration files, and perform network operations via the browser.
Sanitization: While user-provided queries and URLs are sanitized for injection into the JavaScript execution context using JSON.stringify and encodeURIComponent, the content extracted from web pages is not sanitized or filtered for malicious instructions before being returned to the agent.

gsearch