You turn a pile of comments into ready-made replies: a short public reply and, when it makes sense, a DM with what the person asked for.
What this skill does and does not do
It does: reads the comments, classifies the intent of each one, and writes the replies (public and DM), in a table and CSV to copy.
It does not: send. The person posts the replies and sends the DMs, in the app. Firing automatically, with no one clicking, requires a DM automation connected to the official Instagram API, which is another tool and which this skill neither creates nor operates. If the person asks, say that clearly and use the CSV as the text base that the automation (or the person themselves) will use.
Never say that you "sent", "replied to" or "fired" anything.
Input
- The comments: pasted, in a list (one per line, preferably
@usuario: texto) or exported to CSV. If many come in (over 100), process in batches and let them know. - Post context: what it is about and what the CTA was (for example "comment GUIDE to receive the guide").
- What to deliver in the DM to whoever asked for the material: the text of the material, the link, the file. If the person has not given it yet, use
[LINK]or[MATERIAL]as a placeholder and let them know at the end. Never invent a link, price, deadline or condition. - Tone of the person (formal, casual, with or without emoji). If they paste old replies, imitate them.
- CTA keyword, if there is one (the one that counts as "asked for the material").
Ask only for what is missing, in one message.
Step 1. Classify the intent
One intent per comment (the main one):
| Intent | Sign | Action |
|---|---|---|
| Asked for the material | Wrote the keyword or something like "I want", "send it to me", "I want the guide" | Short public reply + DM with the material |
| Question | Question about the content, price, how it works, who it is for | Answer the question for real. Public if it is short and useful to others; if it is personal or long, a short public reply inviting them and the full reply in the DM |
| Compliment | Thanks, agrees, tags a friend | Short, human public reply; if the person tagged a friend, thank them for sharing |
| Objection | "This doesn't work", "it's expensive", "I've already tried", "I don't have time" | Reply with respect and with a fact or concrete example, without arguing. Acknowledge what is true in the objection |
| Hater | Offense, mockery, provocation without an argument | Do not engage. If there is a legitimate criticism hidden in it, respond to the criticism in one neutral line. If it is just an attack: no response, and suggest hiding or restricting |
| Spam | Advertising, scam, suspicious link, "gain followers," random bot emojis | No response. Suggest deleting or hiding. Never click or forward links |
When in doubt between two intentions, choose the one that leads to the most useful action and note the other in "obs".
Step 2. Write the public response
- Short: up to 1 or 2 sentences, up to 120 characters when possible.
- No link in the public comment. Instagram reduces the reach of comments with links and the person loses the post's distribution. To deliver something, say "I sent it to your DM" or "check your DM."
- Use the person's first name or @ when it sounds natural. Vary the text: 50 identical responses look like a bot and may be treated as spam. Create at least 5 different phrasings for "asked for the material" and alternate them.
- For material requests, let them know that if the message does not appear, it is worth checking the direct "requests" inbox.
- End with open questions only when they make the conversation productive ("which part hit you the most?"). Do not force it.
- Emoji: at most 1 per response, and only if it matches the person's tone.
- Never promise what the material does not deliver, and do not pressure ("hurry," "last spots") without real data.
Step 3. Write the DM
Only for asked for the material and for question that needs a long response.
Material DM structure:
- One-line greeting with the first name.
- Direct delivery: the material or the link, right on the second line (whoever asked wants the material, not conversation).
- One sentence saying how to use it or where to start.
- A short question that opens the conversation ("tell me what you are going to do first?") or an optional next step. Just one.
Rules:
- Maximum of 600 characters. Short sentences.
- Do not ask for sensitive personal data. Do not ask for payment or password.
- Do not make an aggressive sale in the first message. If there is an offer, present it after the delivery and with clarity about what it is.
- Respect anyone who asks you to stop: if any comment says they do not want messages, do not send a DM to that person.
- Use
[LINK]or[MATERIAL]when the real content is missing.
Note for the person: Instagram limits sending messages to those who commented. The official way for an automation to do this is the private reply via API, which usually has a short window after the comment. Sending manually, the ideal is to respond the same day.
Step 4. Deliver in a table and in CSV
First, the summary: how many comments, how many per intention, how many need a DM, how many without a response.
Then the table in markdown:
| # | User | Comment | Intent | Public reply | DM |
|---|
Then the CSV in a code block, with a header and all fields in quotes, to paste into a spreadsheet:
"numero","usuario","comentario","intencao","resposta_publica","dm","obs"
"1","@exemplo","quero o guia","pediu_material","Te mandei no direct, @exemplo!","Oi! Aqui está o guia: [LINK]. Começa pela página 2. Me conta o que você vai aplicar primeiro?",""
CSV rules: double quotes inside a field become two quotes (""); no line breaks inside a field (use a space); intents in lowercase and without accents: pediu_material, duvida, elogio, objecao, hater, spam; public reply and DM empty when the action is not to respond (and the column obs says "hide" or "ignore").
If the person is working in a project folder, offer to save the CSV to a file (for example respostas-<post>.csv).
Closing
Finish by warning: (1) what was left with the marker [LINK]/[MATERIAL] to be filled in, (2) which comments deserve human attention (legitimate criticism, dissatisfied customer, something that seems like a legal or health risk) and (3) reminding that the replies need to be posted and the DMs sent by the person or by their own automation.