content-factory

name	content-factory
description	Design spec for a fully automated Content Factory inside Discord - multi-agent architecture for trending topics, scripts, visuals, and organized output.
trigger	content-factory
category	creative

Content Factory — Discord Multi-Agent Architecture

1. System Overview

[Trend Scanner] → [Topic Selector] → [Script Writer] → [Visual Generator]
      ↓                  ↓                 ↓                ↓
[Database] ←─────────────────────────────────────────────────
      ↓
[Discord Output Channels]

2. Discord Server Structure

Category: CONTENT PIPELINE

Channel	Purpose
#trending-topics	Auto-posted trending subjects
#approved-topics	Manually approved topics for production
#scripts-drafts	Generated scripts for review
#scripts-final	Approved scripts ready for production
#thumbnails	Generated visual assets
#content-queue	Final content ready to publish

Category: INTELLIGENCE

Channel	Purpose
#topic-intel	Trending data, analytics
#performance-metrics	Content performance tracking
#feedback-loop	Human corrections fed back

Category: AGENTS

Channel	Purpose
#agent-log	Agent activity logs
#agent-errors	Errors requiring attention

Category: OPERATIONS

Channel	Purpose
#commands	Bot command input
#daily-brief	Daily content brief from Hermes
#alerts	System notifications

3. Multi-Agent System

Agent	Role	Input	Output
Trend Scanner	Monitor trends	Twitter/X API, Google Trends, Reddit	List of trending topics
Topic Selector	Filter & rank	Trending topics list + criteria	3-5 approved topics
Script Writer	Write scripts	Approved topic	Video script (hook + body + CTA)
Visual Generator	Create assets	Final script	Thumbnail + post images
Quality Controller	Validate output	Script + visuals	PASS/FAIL + feedback
Publisher	Format & post	Final content	Discord message + metadata

4. Workflow Pipeline

1. TREND SCAN (every 6h)
   → Fetch: Twitter API, Google Trends, Reddit
   → Filter: relevance score > threshold
   → Output: trending-topics channel

2. TOPIC APPROVAL (automatic + manual)
   → Hermes reviews and ranks
   → User approves or rejects
   → Output: approved-topics channel

3. SCRIPT GENERATION (triggered on approval)
   → Script Writer agent picks topic
   → Generates: hook, main content, CTA
   → Output: scripts-drafts channel

4. QUALITY CHECK (automatic)
   → Quality Controller reviews
   → PASS → scripts-final
   → FAIL → feedback-loop + retry

5. VISUAL GENERATION (triggered on script approval)
   → Visual Generator creates thumbnail
   → Creates social media assets
   → Output: thumbnails channel

6. PUBLISH READY (automatic)
   → Publisher formats for each platform
   → Output: content-queue channel

5. Automation Stack

Tool	Purpose
Hermes Agent	Orchestrator, runs all agents
Discord Bot	Channel management, posting
Twitter/X API	Trend monitoring
Google Trends API	Trend data
DALL-E / Flux	Thumbnail generation
Claude/GPT	Script writing
n8n / Make	Workflow automation glue
Redis	Queue management

6. Leverage & Scalability

• Agents run in parallel — multiple scripts simultaneously
• Queue-based processing — no blocking
• Templates — script structure reused, only content changes
• Modular agents — add new formats without rebuilding
• Human-in-the-loop only at approval gates

7. Quality Control

• Automatic scoring: engagement prediction, completeness check
• Human review: scripts and thumbnails must be approved
• Feedback loop: corrections improve next outputs
• Performance tracking: metrics per topic, per format, per agent

8. Common Mistakes to Avoid

1. Skipping the approval gate — fully automated content loses quality
2. Too many agents — start with 3, add only when necessary
3. No feedback loop — agents repeat mistakes without corrections
4. Monolithic design — keep agents small and replaceable
5. Ignoring metrics — track what works, kill what doesn't