Skip to content
AgentCustomer Service & Support

Comment and DM Reply Agent

Agent kit that triages your Instagram, Facebook, TikTok and YouTube comments and DMs, drafts replies in your voice, spots scam DMs and flags what a person must handle.

Updated Oct 30, 2026Marketplace Founder

About Comment and DM Reply Agent

A busy post brings in a mix of messages. There are questions, orders, a complaint, spam, maybe a fake "Meta support" notice, and now and then something serious. Answering them in the wrong order, or with the wrong words in public, costs sales and trust.

Paste in a batch. The agent sorts every message into one of 10 categories and gives it a priority: P1 now, P2 today, P3 later. It chooses the action: reply publicly, move to DM, hide, report or escalate. Then it drafts each reply from your voice card, short in public and private when personal details are involved. It never promises refunds, discounts or dates beyond your policy, and never puts customer details in public. Scam DMs get no reply and a plain explanation of why they're fake. Injuries, allergy reactions, legal threats and press go to a person, with notes on what to check first. Questions that keep coming up become saved replies.

Inside: the agent instructions, triage rules with escalation triggers and scam patterns, a reply playbook with a voice card template and platform notes, conversation starters, a Claude Code subagent and a worked example with ten real-world messages.

For small businesses, creators and social media managers.

Limitations: it drafts only and doesn't connect to any platform. Replies are only as accurate as the facts you give it. It isn't legal or medical advice.

Scanned: no issues found

Release v1.0.0 reviewed by AgentLoomX™ · 9 files · 16.1 KB

Signed release

This package can

Nothing special: the scan found no shell commands, network calls, hooks or access to secrets.

What's inside (9 files)
  • comment-dm-reply-agent/CHANGELOG.md
  • comment-dm-reply-agent/claude-code/agents/comment-dm-reply-agent.md
  • comment-dm-reply-agent/examples/example-1.md
  • comment-dm-reply-agent/instructions.md
  • comment-dm-reply-agent/knowledge/reply-playbook.md
  • comment-dm-reply-agent/knowledge/triage-rules.md
  • comment-dm-reply-agent/LICENSE.txt
  • comment-dm-reply-agent/README.md
  • comment-dm-reply-agent/starters.md
Read the README comment-dm-reply-agent/README.md
# Comment and DM Reply Agent

An agent kit that sorts your comments and direct messages from Instagram, Facebook, TikTok and YouTube, and drafts the replies. Paste in a batch. You get back:
- a triage table: category, priority (P1 now, P2 today, P3 later), the action and why;
- a reply for each message in your voice, public or private as fits;
- a list of what a person must handle (injuries, legal threats, press, anything risky);
- saved-reply candidates for questions that keep coming up.

It spots scam DMs, such as fake "Meta support" copyright notices, so nobody clicks the link. It never promises refunds or fixes beyond your policy.

Made by AgentLoomX (https://agentloomx.com). Version 1.0.0.

## Who it's for

- Small businesses that handle their own social inboxes.
- Creators with more comments than time.
- Social media managers covering community for clients.

## What's inside

```
comment-dm-reply-agent/
  instructions.md                 the agent's instructions (system prompt)
  knowledge/triage-rules.md       10 categories, priorities, escalation triggers, hide/delete/report, scam DM patterns
  knowledge/reply-playbook.md     voice card template, reply patterns, platform notes, phrases to avoid
  starters.md                     conversation starters
  claude-code/agents/comment-dm-reply-agent.md   Claude Code subagent version
  examples/example-1.md           worked example: ten messages for a bakery
  README.md, CHANGELOG.md, LICENSE.txt
```

## Setup

### ChatGPT (GP

…

The full file is included with your purchase.

SHA-256 ade0c7fd1e712048b469c44e7a62cbb5940dbf63316f4b613e0dcb82fdcf3b38

Signed with Ed25519 key a9c62a7921065aa9. Public keys

Scans catch known risky patterns; they are not a guarantee. Report anything that looks wrong to support.

Highlights

  • Triage table: category, priority, action
  • Replies in your voice, public or private
  • Spots fake support and phishing DMs
  • Escalates injuries, legal threats, press
  • Never promises beyond your policy
  • Turns repeat questions into saved replies
  • social media
  • community management
  • instagram
  • facebook
  • tiktok
  • customer service
  • replies