The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references industry-standard SDKs and APIs from well-known media and AI vendors, including Live2D (Cubism SDK), Pixiv (VRM), and specialized TTS providers like Cartesia and Fish Audio. These references are used for documentation and integration purposes within the streaming domain.
[COMMAND_EXECUTION]: The provided code snippets demonstrate the use of the obs-websocket-js library for automating OBS Studio and the three-vrm library for avatar rendering. These operations are performed via established WebSocket protocols and official SDKs, following local control patterns.
[PROMPT_INJECTION]: While the skill interacts with untrusted live chat data, it features an extensive Content Safety reference (references/content-safety.md) that defines robust defenses against indirect prompt injection. This includes the use of dedicated LLM classifiers, structured input channels, and output moderation APIs. The presence of these instructions significantly reduces the risk of successful injection attacks.
[DATA_EXFILTRATION]: No evidence of hardcoded credentials or data exfiltration was found. The skill explicitly advises against hardcoding sensitive information and recommends the use of environment variables and secure OAuth flows for platform integrations.

aether