The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill allows the agent to generate and render functional HTML and JavaScript code (Alpine.js) within an iframe sandbox using tools like render-inline-extension and create-extension. This enables the agent to dynamically create executable components based on its own instructions.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates an environment where data processed by the generated UI can influence the agent's behavior, creating a potential attack surface.
Ingestion points: External data enters the agent context through window.slotContext and external API responses via extensionFetch as described in SKILL.md.
Boundary markers: UI components are isolated within a sandboxed Alpine.js iframe.
Capability inventory: The generated UI can send chat messages (agentNative.chat.send), write application state (agentNative.ui.output), and execute database commands (dbExec).
Sanitization: The skill instructions advise against hardcoding secrets or calling internal framework routes, though no programmatic sanitization of external input is enforced.
[COMMAND_EXECUTION]: The skill exposes the dbQuery and dbExec tools, which allow the agent to execute SQL operations on the application's database tables.

generative-ui