When a video does well, the comment section turns into homework. Some people ask real questions, some say thank you, a few are unhappy, and plenty is spam. Answering all of it in one sitting makes replies sound flat.
The comment-reply-assistant skill for Hermes splits the job in two. A small script sorts an exported comments file (CSV or plain text) into questions, praise, criticism, spam and other. It also flags the comments that need a careful human answer. Then the agent drafts a reply for each row in your voice, using a few replies you have written before as a reference.
Nothing is posted. Every row stays marked draft until you approve it, and anything the agent cannot know (a price, a date, a personal detail) is left as a [YOU: fill in ...] placeholder instead of an invented answer. Critical comments get a calm draft and a warning, and abusive ones get no reply at all.
A tiny example, with four made-up comments:
- "Can you do one on lighting?" becomes a question, with a draft that says yes and leaves the date for you.
- "Great video, I learned a lot!" becomes praise, with a short and specific thank-you.
- "Too long, clickbait title." becomes criticism, marked for extra care.
- "sub4sub follow me" becomes spam, no reply.
The sorting is based on simple keyword rules in English and French, so it will misread irony and mixed messages. Treat it as a first pass, not a judge. It needs no account and no API key, only Python 3.
Skill (MIT): https://github.com/RobinBeraud/hermes-skills/tree/main/content-creation/comment-reply-assistant
All skills: https://github.com/RobinBeraud/hermes-skills
This post was written with AI assistance and reviewed through automated checks.
Top comments (0)