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
- MUST match brand voice (use voice_profile_id)
- MUST create draft (not publish directly) unless role=editor AND Safe Mode off
- MUST escalate sensitive topics (refunds, complaints, legal, billing)
- MUST NOT use generic template responses
- 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)
1---2name: bolta-reply-to-mention3version: 2.0.04description: Draft a reply to a mention, comment, or DM while maintaining brand voice and community presence5category: engagement6roles_allowed: [Viewer, Creator, Editor, Admin]7agent_types: [engagement, custom]8safe_defaults:9 always_draft_for_approval: true10 escalate_sensitive_topics: true11tools_required:12 - bolta.get_voice_profile13 - bolta.create_reply_draft14inputs_schema:15 type: object16 required: [mention_id, reply_text]17 properties:18 mention_id: { type: string, description: "Mention UUID to reply to" }19 reply_text: { type: string, description: "Draft reply content" }20 tone: { type: string, enum: [helpful, empathetic, professional], description: "Reply tone" }21 escalate: { type: boolean, description: "Flag for human review", default: false }22 voice_profile_id: { type: string, description: "Voice profile to match" }23outputs_schema:24 type: object25 properties:26 success: { type: boolean }27 draft_reply_id: { type: string }28 status: { type: string, enum: [draft] }29 requires_approval: { type: boolean }30 escalated: { type: boolean }31organization: bolta.ai32author: Bolta Team33---3435## Goal36Draft a reply to a mention, comment, or DM while staying on-brand and maintaining community presence.3738## Which Agents Use This39- **engagement** — Respond to common questions, compliments, and community interactions40- **moderator** — De-escalate complaints with empathetic responses41- **custom** — Any agent handling community engagement4243## Hard Rules441. MUST match brand voice (use voice_profile_id)452. MUST create draft (not publish directly) unless role=editor AND Safe Mode off463. MUST escalate sensitive topics (refunds, complaints, legal, billing)474. MUST NOT use generic template responses485. Tone should match context (complaint = empathetic, question = helpful, compliment = warm)4950## Steps5152### 1. Validate input53- Verify mention_id exists54- Verify reply_text is non-empty55- Load voice_profile if provided5657### 2. Check for sensitive topics58- If content mentions: refund, billing, legal, broken, angry, lawsuit59 - Set escalate=true automatically60 - Flag for human review6162### 3. Create reply draft63- Create draft reply linked to mention64- Apply voice profile for brand consistency65- Apply tone (helpful, empathetic, professional)66- Mark requires_approval=true6768### 4. Return draft confirmation69- Return draft_reply_id, status, escalated flag7071## Output72```json73{74 "success": true,75 "draft_reply_id": "uuid",76 "status": "draft",77 "requires_approval": true,78 "escalated": false79}80```8182## Failure Handling83- If mention_id not found: return error "Mention not found"84- If reply_text empty: return error "Reply text required"85- If voice_profile_id invalid: use workspace default voice8687## Example Usage8889### Scenario 1: Helpful response to question90```json91{92 "mention_id": "uuid",93 "reply_text": "Hi! You can reset your password from Settings → Security. Let me know if you need help!",94 "tone": "helpful",95 "voice_profile_id": "uuid"96}97```98**Result:** Draft reply created for approval99100### Scenario 2: Escalate complaint101```json102{103 "mention_id": "uuid",104 "reply_text": "We're sorry to hear about your experience. Our team is looking into this.",105 "tone": "empathetic",106 "escalate": true107}108```109**Result:** Draft reply created AND flagged for human review (complaint escalation)
boltaai