The Agent Skills Directory

[COMMAND_EXECUTION]: The build process and 'Speedrun Prompt' involve the execution of shell commands such as 'npm create vite' for project scaffolding, 'npm install' for dependency management, and Python scripts for image processing and screenshot comparison. These operations are standard practices for the intended technical use case of the skill.
[EXTERNAL_DOWNLOADS]: The skill directs the agent to download assets (logos, fonts) from user-specified target websites and to install community-standard libraries from the npm registry (e.g., 'tailwindcss', 'framer-motion'). These activities are central to the primary purpose of website cloning.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it navigates to and processes data from arbitrary external URLs. An adversary could embed instructions within a website's HTML or metadata to influence the agent during the extraction phase. \n * Ingestion points: Data is ingested via 'navigate' and 'javascript_tool' calls that read content from external target websites. \n * Boundary markers: The provided scripts do not include explicit markers or instructions to isolate or ignore potentially adversarial text within the crawled content. \n * Capability inventory: The agent possesses the capability to write files via 'bash' and execute local commands via 'npm'. \n * Sanitization: Extracted content like 'innerText' or 'outerHTML' is used directly in the cloning workflow without validation or sanitization layers.

clone-anywebsite