The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates reading game state data into the agent's context through browser inspection methods. This creates a surface where game-originated content could influence the agent's behavior.
Ingestion points: Methods such as window.nqtr.getRooms(), window.nqtr.getActivities(), and window.nqtr.getQuests() in SKILL.md are used to read domain objects from the browser context.
Boundary markers: No specific delimiters or instructions to treat ingested data as non-executable text are provided.
Capability inventory: The agent exerts control over the game state through actions like goToRoom, increaseTime, and runActivity as documented in SKILL.md.
Sanitization: There is no evidence of sanitization for strings or data structures ingested from the game engine's environment.
[DYNAMIC_EXECUTION]: The skill is designed around the agent using browser evaluation to interact with and inspect the game's testing API.
Evidence: The "Drive actions" section of SKILL.md provides several JavaScript snippets intended for execution within the browser's global scope via the window.nqtr object.

nqtr-testing