Feedback Loop Wiring Summary

Date: 2026-02-05
Status: ✅ Complete - All agents in brainforge-vault are now wired to the automated feedback loop


What Was Done

1. Created Agent Registry

File: AGENT_REGISTRY.md

Complete inventory of all 12 agents in brainforge-vault, organized by category:

  • Content Generation (3 agents)
  • Campaign Planning (2 agents)
  • Outreach (4 agents)
  • Analysis (2 agents)
  • Operations (1 agent)

2. Created Feedback Prompt Templates

Location: feedback-prompts/

Created 11 feedback prompt templates (one already existed for ticket creation):

TemplateAgent
design-ready-copy-agent.mdDesign-Ready Copy Agent
campaign-brief-agent.mdCampaign Brief Intake Agent
campaign-post-agent.mdCampaign Post Agent (via CC content system)
message-sequence-agent.mdMessage Sequence Agent
slack-deployment-agent.mdSlack Deployment Agent
event-follow-up-agent.mdEvent Follow-Up Agent
linkedin-sequence-agent.mdLinkedIn Sequence Agent
icp-analysis-agent.mdICP Analysis Agent
metrics-teardown-agent.mdMetrics Teardown Agent
ticket-event-follow-up-agent.mdTicket Event Follow-Up Agent
vp-partnerships-agent.mdVP Partnerships Agent
ticket-creation-agent.mdTicket Creation Agent (already existed)

3. Updated Core Documentation

Files Updated:

  • AGENT_FEEDBACK_LOOP.md - Added “Covered Agents” section referencing registry
  • All agents now documented as part of the feedback loop system

How It Works Now

Every time you run any agent in brainforge-vault:

  1. Agent runs → Generates output
  2. System prompts → You get agent-specific feedback questions (2-3 min)
  3. Auto-log → Run logged to RUN_LOG.md with metadata
  4. Pattern analysis → System analyzes patterns (thinking → summary)
  5. Impact summary → Shows time saved, quality trends, knowledge impact
  6. PR suggestions → When patterns reach MEDIUM confidence, suggests agent improvements

Agents Now Covered

Design-Ready Copy Agent - Marketing collateral copy
Campaign Brief Intake Agent - Execution-ready campaign briefs
Ticket Creation Agent - Linear tickets from requests
Message Sequence Agent - Multi-step outreach sequences
Campaign Post Agent - LinkedIn campaign posts
Slack Deployment Agent - Deployment win messages
Event Follow-Up Agent - Post-event outreach
LinkedIn Sequence Agent - LinkedIn-specific sequences
Ticket Event Follow-Up Agent - Follow-up automation
ICP Analysis Agent - Attendee list analysis
Metrics Teardown Agent - Metrics/KPI analysis
VP Partnerships Agent - Partnership outreach


Next Steps

  1. Test the system - Run any agent and provide feedback
  2. Monitor patterns - After 5+ runs per agent, patterns will emerge
  3. Create PRs - When patterns reach MEDIUM confidence, update agent PRDs
  4. Roll out to other repos - If this works well, expand to other repositories


Status: Ready for testing. All agents are wired and ready to learn from feedback.