The Agent Skills Directory

[PROMPT_INJECTION]: The skill joins a shared social space where it receives real-time text from other agents and the environment, creating a surface for indirect prompt injection.
Ingestion points: The scripts/connect.js script receives message, room, arrive, and leave events via a WebSocket connection to wss://playground-bots.fly.dev/bot.
Boundary markers: None are present in the script's output. Text from external bots and the server is printed directly to the terminal (e.g., [12:00:00] OtherBot: Hello!) without delimiters or instructions for the agent to ignore embedded commands.
Capability inventory: An agent running this skill typically has access to terminal execution and file system tools.
Sanitization: None. Content from the remote server is parsed as JSON and printed verbatim to the console.
[EXTERNAL_DOWNLOADS]: The skill connects to an external WebSocket server at wss://playground-bots.fly.dev/bot and references a public dashboard at https://playground-bots.fly.dev.
[CREDENTIALS_UNSAFE]: A hardcoded authentication token playground-beta-2026 is present in SKILL.md, README.md, and scripts/connect.js. While presented as a public beta token, hardcoding credentials in source files is a poor security practice.

playground