Brainforge Knowledge

ARCHITECTURE

4 min read

Agent Architecture: Workers & Workflows

Purpose: Unified structure for all agents in brainforge-vault. Each agent is a specialized worker that can be coordinated in multi-agent workflows.
Last Updated: 2026-02-05

Structure Overview

gtm/agents/
├── README.md                          # Overview & quick start
├── ARCHITECTURE.md                    # This file (structure & design)
├── AGENT_REGISTRY.md                  # Complete inventory
│
├── workers/                           # Individual agents (specialized jobs)
│   ├── content-generation/
│   │   ├── design-ready-copy/
│   │   │   ├── PRD.md
│   │   │   ├── feedback-prompt.md
│   │   │   └── examples/
│   │   ├── campaign-post/
│   │   │   ├── PRD.md
│   │   │   ├── feedback-prompt.md
│   │   │   └── examples/
│   │   └── slack-deployment/
│   │       ├── PRD.md
│   │       ├── feedback-prompt.md
│   │       └── examples/
│   │
│   ├── campaign-planning/
│   │   ├── campaign-brief-intake/
│   │   │   ├── PRD.md
│   │   │   ├── feedback-prompt.md
│   │   │   └── examples/
│   │   └── message-sequence/
│   │       ├── PRD.md
│   │       ├── feedback-prompt.md
│   │       ├── playbooks/
│   │       └── examples/
│   │
│   ├── outreach/
│   │   ├── event-follow-up/
│   │   ├── linkedin-sequence/
│   │   ├── ticket-event-follow-up/
│   │   └── vp-partnerships/
│   │
│   ├── analysis/
│   │   ├── icp-analysis/
│   │   └── metrics-teardown/
│   │
│   └── operations/
│       └── ticket-creation/
│
├── workflows/                         # Multi-agent coordination
│   ├── campaign-launch-workflow.md    # Campaign brief → copy → posts → sequences
│   ├── event-to-meeting-workflow.md   # Event follow-up → ICP analysis → sequences
│   └── deployment-workflow.md         # Agent deployment → Slack message → feedback
│
├── shared/                            # Shared infrastructure
│   ├── feedback-loop/
│   │   ├── AGENT_FEEDBACK_LOOP.md
│   │   ├── FEEDBACK_LOOP_PROCESS.md
│   │   └── RUN_LOG.md
│   ├── context-graph/
│   │   ├── CONTEXT_GRAPH_APPROACH.md
│   │   └── PR_CONTEXT_GRAPH_CHECKLIST.md
│   └── patterns/
│       └── PATTERNS.md
│
└── examples/                          # Example runs & outputs
    ├── deployments/                   # Slack deployment messages
    └── [organized by worker]          # Example outputs per worker

Design Principles

1. Workers = Specialized Jobs

Each agent is a specialized worker that does one thing well:

  • Input: Well-defined inputs (campaign brief, ticket request, event data)
  • Output: Well-defined outputs (copy doc, ticket file, sequence)
  • Feedback: Every run prompts for feedback → logs → patterns

2. Workflows = Multi-Agent Coordination

Workflows orchestrate multiple workers to complete complex tasks:

Example: Campaign Launch Workflow

1. Campaign Brief Intake Worker → Creates brief
2. Design-Ready Copy Worker → Generates 2-pager
3. Campaign Post Worker → Drafts LinkedIn posts
4. Message Sequence Worker → Creates outreach sequences
5. Slack Deployment Worker → Announces deployment

3. Shared Infrastructure

Common systems that all workers use:

  • Feedback Loop - Every worker run → feedback → log → patterns
  • Context Graph - Process knowledge & traceability
  • Patterns - Learned behaviors from runs

Worker Structure

Each worker lives in workers/[category]/[worker-name]/:

workers/content-generation/design-ready-copy/
├── PRD.md                    # Product requirements (what it does, inputs, outputs)
├── feedback-prompt.md        # Feedback template (from feedback-prompts/)
├── examples/                 # Example runs/outputs
│   └── insurance-broker-2pager.md
└── README.md                # Quick reference (optional)

Required files:

  • PRD.md - What the worker does, inputs, outputs, rules
  • feedback-prompt.md - Feedback template (links to shared/feedback-loop/)

Optional files:

  • examples/ - Example runs/outputs
  • README.md - Quick reference

Workflow Structure

Each workflow lives in workflows/:

workflows/campaign-launch-workflow.md
├── Overview (what it does)
├── Workers Used (which workers coordinate)
├── Flow Diagram (step-by-step)
├── Inputs/Outputs
└── Examples

Example Workflow:

# Campaign Launch Workflow
 
**Workers:**
1. Campaign Brief Intake Worker
2. Design-Ready Copy Worker
3. Campaign Post Worker
4. Message Sequence Worker
 
**Flow:**
Brief → Copy → Posts → Sequences → Deployment
 
**Input:** Campaign idea/requirements
**Output:** Complete campaign ready to launch

Migration Plan

  1. Create new structure (workers/, workflows/, shared/)
  2. Move existing agents to workers/[category]/[agent-name]/
  3. Create workflow docs for common multi-agent flows
  4. Update references (registry, docs, feedback prompts)
  5. Archive old structure (keep for reference, mark deprecated)

Benefits

✅ Clear organization - All agents in one place, organized by function
✅ Multi-agent ready - Workflows show how workers coordinate
✅ Scalable - Easy to add new workers or workflows
✅ Discoverable - Clear structure makes it easy to find agents
✅ Maintainable - Shared infrastructure in one place

Graph View