The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes UI element trees and labels from external TV applications, creating a surface for indirect instructions.- Ingestion points: The describe tool returns element labels, selected states, and text content from the active TV application screen (SKILL.md).- Boundary markers: The skill does not provide instructions to the agent to treat UI content as untrusted data or use specific delimiters to separate application data from instructions (SKILL.md).- Capability inventory: The agent has access to keyboard for text input, open-url for deep linking, tv-remote for navigation, and debugger-evaluate for JavaScript execution in the target runtime (SKILL.md).- Sanitization: No sanitization, filtering, or validation logic is specified for the UI data ingested by the describe tool before the agent processes it (SKILL.md).- [COMMAND_EXECUTION]: The skill utilizes host-side CLI tools to manage and interact with TV devices.- The agent is instructed to use adb for Android TV (leanback) interactions, xcrun simctl for Apple TV simulator commands, and the vega and vsm CLI tools for Amazon Fire TV (Vega) management (SKILL.md).- [DYNAMIC_EXECUTION]: The skill includes tools to evaluate code within the target application's runtime for debugging.- The debugger-evaluate tool allows the agent to execute JavaScript directly in the TV application's runtime on Vega devices (SKILL.md).

argent-tv-interact