Chad Yi Orchestrate

Pattern 1: Single Agent Assignment

User request → Match to agent → Spawn → Review output

Example:

• Task: Write RE:UNITE Chapter 13
• Agent: Escritor
• Action: Spawn with escritor-novel skill
• Review: Read output, provide feedback

Pattern 2: Multi-Agent Coordination

Complex task → Decompose → Dispatch → Integrate → Verify

Example:

• Task: Launch new product (B1-B10)
• Decompose: Website (B1), SEO (B3), Content (B8)
• Dispatch: 3 agents in parallel
• Integrate: Review all outputs
• Verify: Check consistency across deliverables

Pattern 3: Escalation Handler

Agent blocked → Alert received → Diagnose → Resolve/Delegate

Example:

• Alert: Quanta blocked (needs OANDA credentials)
• Diagnose: Missing API access
• Options:
  • Provide credentials if available
  • Create task for Caleb to obtain
  • Reprioritize work

Agent Roster & Responsibilities

Decision Framework

When to Spawn an Agent

✅ Spawn when:

• Work is independent and parallelizable
• Requires specialized skills
• Caleb needs to focus on other things
• Task is well-defined with clear output

❌ Don't spawn when:

• Real-time coordination needed
• Task requires frequent clarification
• Decision authority needed mid-task
• Simpler to do directly

When to Escalate to Caleb

• Blockers requiring credentials/access
• Strategic decisions about priorities
• Budget/resource allocation
• Creative direction for RE:UNITE
• Agent conflicts or issues

Communication Protocol

Agent → CHAD_YI

Location: /agents/[name]/outbox/
Format: [task-id]-[status].json

CHAD_YI → Agent

Location: /agents/[name]/inbox/
Format: [task-id]-directive.md

Broadcast (All Agents)

Location: /agents/message-bus/broadcast/
Format: [timestamp]-[topic].md

Daily Workflow

Morning (08:00)

1. Check dashboard data.json
2. Review overnight agent reports
3. Check urgent deadlines
4. Clear any blockers from previous day

Throughout Day

1. Respond to agent outbox messages
2. Spawn agents for new work
3. Review completed agent outputs
4. Update task statuses

Evening

1. Review day's progress
2. Plan next day's agent assignments
3. Ensure no urgent items left unresolved
4. Update MEMORY.md with decisions

Task Assignment Template

When assigning work to an agent:

# Task Assignment: [Agent Name]

**Task ID:** [A-X-Y]  
**Priority:** High/Medium/Low  
**Deadline:** YYYY-MM-DD

## Objective
[Clear, specific goal]

## Context
[Background information needed]

## Success Criteria
- [ ] Deliverable 1
- [ ] Deliverable 2

## Constraints
- [ ] Don't touch [specific files]
- [ ] Must coordinate with [other agent]

## Escalation
Contact CHAD_YI if:
- Blocked for >4 hours
- Scope creep discovered
- Dependencies unclear

Issue Resolution

Blocker Types & Responses

Metrics to Track

• Tasks completed per agent per week
• Average time from assignment to completion
• Blocker frequency by agent
• Quality of agent outputs (user satisfaction)
• System uptime (dashboard freshness)

Success Criteria

CHAD_YI is successful when:

• Agents work independently with minimal intervention
• Blockers are resolved within 24 hours
• Dashboard accurately reflects reality
• Caleb is unblocked for strategic work
• Projects progress without micromanagement