bolta-reply-to-mention

name	bolta.reply_to_mention
version	2.0.0
description	Draft a reply to a mention, comment, or DM while maintaining brand voice and community presence
category	engagement
roles_allowed	["Viewer","Creator","Editor","Admin"]
agent_types	["engagement","custom"]
safe_defaults	{"always_draft_for_approval":true,"escalate_sensitive_topics":true}
tools_required	["bolta.get_voice_profile","bolta.create_reply_draft"]
inputs_schema	{"type":"object","required":["mention_id","reply_text"],"properties":{"mention_id":{"type":"string","description":"Mention UUID to reply to"},"reply_text":{"type":"string","description":"Draft reply content"},"tone":{"type":"string","enum":["helpful","empathetic","professional"],"description":"Reply tone"},"escalate":{"type":"boolean","description":"Flag for human review","default":false},"voice_profile_id":{"type":"string","description":"Voice profile to match"}}}
outputs_schema	{"type":"object","properties":{"success":{"type":"boolean"},"draft_reply_id":{"type":"string"},"status":{"type":"string","enum":["draft"]},"requires_approval":{"type":"boolean"},"escalated":{"type":"boolean"}}}
organization	bolta.ai
author	Bolta Team

Goal

Draft a reply to a mention, comment, or DM while staying on-brand and maintaining community presence.

Which Agents Use This

• engagement — Respond to common questions, compliments, and community interactions
• moderator — De-escalate complaints with empathetic responses
• custom — Any agent handling community engagement

Hard Rules

1. MUST match brand voice (use voice_profile_id)
2. MUST create draft (not publish directly) unless role=editor AND Safe Mode off
3. MUST escalate sensitive topics (refunds, complaints, legal, billing)
4. MUST NOT use generic template responses
5. Tone should match context (complaint = empathetic, question = helpful, compliment = warm)

Steps

1. Validate input

• Verify mention_id exists
• Verify reply_text is non-empty
• Load voice_profile if provided

2. Check for sensitive topics

• If content mentions: refund, billing, legal, broken, angry, lawsuit
  • Set escalate=true automatically
  • Flag for human review

3. Create reply draft

• Create draft reply linked to mention
• Apply voice profile for brand consistency
• Apply tone (helpful, empathetic, professional)
• Mark requires_approval=true

4. Return draft confirmation

• Return draft_reply_id, status, escalated flag

Output

{
  "success": true,
  "draft_reply_id": "uuid",
  "status": "draft",
  "requires_approval": true,
  "escalated": false
}

Failure Handling

• If mention_id not found: return error "Mention not found"
• If reply_text empty: return error "Reply text required"
• If voice_profile_id invalid: use workspace default voice

Example Usage

Scenario 1: Helpful response to question

{
  "mention_id": "uuid",
  "reply_text": "Hi! You can reset your password from Settings → Security. Let me know if you need help!",
  "tone": "helpful",
  "voice_profile_id": "uuid"
}

Result: Draft reply created for approval

Scenario 2: Escalate complaint

{
  "mention_id": "uuid",
  "reply_text": "We're sorry to hear about your experience. Our team is looking into this.",
  "tone": "empathetic",
  "escalate": true
}

Result: Draft reply created AND flagged for human review (complaint escalation)