The Agent Skills Directory

[PROMPT_INJECTION]: The WebsiteToRedesign workflow retrieves and processes content from user-provided URLs, creating a surface for indirect prompt injection. A malicious website could include hidden instructions to manipulate the agent during the critique or design brief phases. \n
Ingestion points: Workflows/WebsiteToRedesign.md (captures HTML from external URLs). \n
Boundary markers: Absent; content from the URL is integrated into the redesign brief without explicit delimiters. \n
Capability inventory: The skill has the ability to write files, execute deployment commands, and drive a browser session. \n
Sanitization: No sanitization or filtering of the external HTML content is implemented. \n- [COMMAND_EXECUTION]: The skill orchestrates multiple development and deployment CLI tools (e.g., wrangler, vercel, netlify, gh, aws, unzip, and interceptor) to perform design and integration tasks. These represent the primary interface with the host system. \n- [SAFE]: The DeployDesign workflow includes a security-positive step that uses ripgrep (rg) to scan the build output directory for sensitive patterns such as API keys and private keys before any deployment is finalized.

Webdesign