GTM Agents: Workers & Workflows

Purpose: Unified system for all GTM agents (workers) and multi-agent workflows
Status: Active - All workers wired to feedback loop
Last Updated: 2026-02-05


πŸš€ Quick Start

New here? Start with:

  1. Architecture - Structure design (5 min read)
  2. Agent Registry - Complete worker inventory
  3. Pick a workflow: Campaign Launch or Event to Meeting

Want to run a worker?

  • Find it in Agent Registry
  • Read the worker’s PRD.md for inputs/outputs
  • Run it β†’ System prompts for feedback β†’ Logs run β†’ Analyzes patterns

πŸ“‚ Structure

gtm/agents/
β”œβ”€β”€ README.md                    # This file (overview)
β”œβ”€β”€ ARCHITECTURE.md              # Structure design
β”œβ”€β”€ AGENT_REGISTRY.md            # Complete worker inventory
β”‚
β”œβ”€β”€ workers/                     # Individual workers (specialized jobs)
β”‚   β”œβ”€β”€ content-generation/      # Copy, posts, deployment messages
β”‚   β”œβ”€β”€ campaign-planning/       # Briefs, sequences
β”‚   β”œβ”€β”€ outreach/                # Follow-ups, LinkedIn, partnerships
β”‚   β”œβ”€β”€ analysis/                 # ICP analysis, metrics
β”‚   └── operations/              # Ticket creation
β”‚
β”œβ”€β”€ workflows/                   # Multi-worker coordination
β”‚   β”œβ”€β”€ campaign-launch-workflow.md
β”‚   β”œβ”€β”€ event-to-meeting-workflow.md
β”‚   └── deployment-workflow.md
β”‚
β”œβ”€β”€ shared/                      # Shared infrastructure
β”‚   β”œβ”€β”€ feedback-loop/           # Feedback system, run logs
β”‚   β”œβ”€β”€ context-graph/          # Process knowledge
β”‚   └── patterns/               # Learned patterns
β”‚
└── examples/                    # Example runs/outputs

🎯 Workers (Individual Agents)

Each worker is a specialized job that does one thing well. See Agent Registry for complete list.

Content Generation

  • Design-Ready Copy - Marketing collateral (one-pagers, 2-pagers)
  • Campaign Post - LinkedIn posts via CC content system
  • Slack Deployment - Deployment win messages

Campaign Planning

  • Campaign Brief Intake - Execution-ready campaign briefs
  • Message Sequence - Multi-step outreach sequences

Outreach

  • Event Follow-Up - Post-event sequences
  • LinkedIn Sequence - LinkedIn-specific sequences
  • Ticket Event Follow-Up - Follow-up automation
  • VP Partnerships - Partnership outreach

Analysis

  • ICP Analysis - Attendee list analysis
  • Metrics Teardown - Metrics/KPI analysis

Operations

  • Ticket Creation - Linear tickets from requests

πŸ”„ Workflows (Multi-Worker Coordination)

Workflows orchestrate multiple workers to complete complex tasks:

Campaign Launch Workflow

Workers: Brief Intake β†’ Design-Ready Copy β†’ Campaign Post β†’ Message Sequence β†’ Slack Deployment
Input: Campaign idea
Output: Complete campaign ready to launch
See workflow β†’

Event to Meeting Workflow

Workers: ICP Analysis β†’ Event Follow-Up β†’ Message Sequence
Input: Event attendee list
Output: Prioritized sequences ready to send
See workflow β†’

Deployment Workflow

Workers: [Any Worker] β†’ Slack Deployment β†’ Feedback Loop
Input: New agent/tool deployed
Output: Team announcement + feedback logged
See workflow β†’


πŸ” Feedback Loop System

Every worker run:

  1. Worker runs β†’ Generates output
  2. System prompts β†’ Structured feedback (2-3 min)
  3. Auto-log β†’ Run logged to shared/feedback-loop/RUN_LOG.md
  4. Pattern analysis β†’ β€œThinking β†’ summary” approach
  5. Impact summary β†’ Time saved, quality trends, knowledge impact
  6. PR suggestions β†’ When patterns reach MEDIUM confidence

See: Feedback Loop System


πŸ“Š Context Graph

The feedback loop builds a context graph that captures:

  • Entities: Workers, workflows, outputs, patterns
  • Actions/Traces: Who ran what, when, with what outcomes
  • Process Patterns: How work actually gets done (not just what exists)

See: Context Graph Approach


βž• Adding a New Worker

  1. Create worker directory: workers/[category]/[worker-name]/
  2. Create PRD.md: Define inputs, outputs, rules
  3. Create feedback-prompt.md: Copy from similar worker
  4. Add to registry: Update AGENT_REGISTRY.md
  5. First run: Prompt for feedback β†’ log β†’ analyze patterns


Status: All workers wired to feedback loop. System ready for multi-worker workflows.