This is a feedback collection tool that triggers after you've finished a LiveAvatar integration, when the user hits friction, or when they explicitly ask to report something. It reads the conversation history to extract what they were building and what went wrong, then asks permission before sending anything to the HeyGen team. The prompt engineering is careful about consent (shows you the payload first, waits for explicit approval) and keeps user-written feedback verbatim instead of paraphrasing it. It's basically instrumented telemetry for an AI coding assistant, which is interesting because it treats developer experience issues as first-class data worth capturing in the moment rather than hoping someone files a GitHub issue later.
npx skills add https://github.com/heygen-com/liveavatar-agent-skills --skill liveavatar-feedback