Infrastructure Adoption - Week 1 Plan

Goal

Make infrastructure deployment HABITUAL for all 22 orchestrators + 32 personas.

We deployed mandatory infrastructure on 2026-02-03. This plan ensures agents actually USE it in their daily work.

What We Deployed

Already deployed:

• ✅ Incremental sync (90-95% fewer API calls)
• ✅ Rate limiting (80% safety margin, zero errors)
• ✅ WPEngine metrics (228 sites, every 2 hours)
• ✅ Sync profiles (10 profiles for all use cases)

Needs manual deployment:

• ⚠️ Memory system Supabase tables (ready, needs Supabase SQL - 10-15 min)
• ⚠️ WPEngine alerts (ready, needs SSL investigation - 40 min)

Total deployment time remaining: 50-55 minutes

---

Day 1: Complete Manual Steps + First Usage

Morning (1-2 hours)

Task 1: Complete Deployment (50-55 min)

Follow these guides in order:

1. Memory System Deployment (10-15 min)

   • Guide: docs/planning/MANUAL-DEPLOY-memory-system-2026-02-03.md
   • Apply 2 SQL migrations in Supabase
   • Test with quick-test script

2. 1Password Authentication (15-20 min)

   • Guide: docs/planning/1password-sa-setup.md
   • Create service account: fp-tech-automation
   • Grant Engineering vault access
   • Update .env file

3. WPEngine Alerts (40 min)

   • Guide: docs/planning/NEXT-STEPS-wpengine-ssl-2026-02-03.md
   • Test SSL filtering (should show <10 alerts)
   • Deploy to cron (every 15 min)
   • Monitor first run in Slack #tech-alerts

Checkpoint: All systems operational ✅

---

Afternoon (2-3 hours)

Task 2: First Real Usage - Demonstrate the Workflow

Pick a REAL task and demonstrate the full mandatory workflow.

Example Task: "Add backup monitoring dashboard to fp-interface"

# Step 1: SEARCH FOR RELATED DECISIONS (MANDATORY)
cd ~/GitHub/flypilot
grep -ril "dashboard" docs/planning/decisions/
grep -ril "backup" docs/planning/decisions/
grep -ril "monitoring" docs/planning/decisions/
grep -ril "wpengine" docs/planning/decisions/

# Step 2: CHECK HANDOFFS
ls handoffs/

# Step 3: RECORD ARCHITECTURAL DECISION
cat > docs/planning/decisions/$(date +%Y-%m-%d)-use-wpengine-backup-status-view.md << 'EOF'
# Decision: Use v_wpengine_backup_status view for dashboard

**Date:** 2026-02-25
**Status:** Active
**Impact:** medium
**Category:** architecture
**Department:** fp

## Decision

Use the existing v_wpengine_backup_status Supabase view for the backup
monitoring dashboard in fp-interface.

## Rationale

View already exists from WPEngine Phase 1. Provides unified backup status
across 228 sites without additional query complexity.

## Alternatives Considered

- **Query wpengine_metrics directly:** More complex query, duplicates view logic
- **Build custom aggregation:** Unnecessary given the view already exists
EOF

# Step 4: Do the Work
# ... implement the feature ...

# Step 5: DOCUMENT IMPORTANT FINDINGS
echo "- $(date +%Y-%m-%d): v_wpengine_backup_status view has 228 sites, updated every 2 hours via Mac Studio cron — tags: wpengine,dashboard,monitoring,backup" >> docs/planning/research-notes.md
echo "- $(date +%Y-%m-%d): WPEngine health views use 0-100 score calculation: backup=40%, ssl=30%, disk=20%, bandwidth=10% — tags: wpengine,monitoring,health-score" >> docs/planning/research-notes.md

# Step 6: Verify Decisions Were Recorded
ls -lt docs/planning/decisions/ | head -5

Document This Example

Create: docs/planning/infrastructure-workflow-example.md

Include:

• Full command sequence from above
• When to record decisions (any choice between 2+ alternatives)
• When to document findings (key facts, non-obvious insights)
• When to search first (ALWAYS before deciding)

This becomes the reference example for all other agents to follow.

---

Day 2-3: Agent Training

Goal: Ensure all orchestrators know about the new workflows

Training Protocol (For Each Orchestrator)

Before Training:

1. Start session in tmux: tmux new -s training-[role]
2. Load their role file: cat docs/planning/agent-roles/[role].md
3. Review their "Before Every Session" checklist

During Training:

1. Pick a real task from their backlog (not a fake task)
2. Walk through the mandatory workflow:
   • Search for related decisions with grep
   • Check handoffs for ongoing context
   • Record decisions made during work as markdown files
   • Document key findings in research notes
3. Observe what worked / what was confusing
4. Note any workflow improvements needed

After Training:

1. Document in training log
2. If issues found, update workflows or documentation
3. Celebrate successful adoption

Recommended Training Schedule

Day 2 Orchestrators:

1. FP-Infrastructure (infrastructure role)

   • Task: "Investigate Mac Studio backup automation"
   • Focus: Decision recording for infrastructure choices

2. OMG-Accounts (client role)

   • Task: "Audit OMG client Toggl time entries"
   • Focus: Using sync profiles instead of custom API calls

3. FP-Research (research role)

   • Task: "Compare LLM context window offerings"
   • Focus: Documenting research findings in notes files

Day 3 Orchestrators:

1. FP-Core (development role)

   • Task: "Refactor sync rate limiting"
   • Focus: Full workflow (search → decide → record → document)

2. LRND-PM (project management role)

   • Task: "Generate weekly project status report"
   • Focus: Using sync profiles for data access

3. All-Accounts (accounts director)

   • Task: "Review client time allocation patterns"
   • Focus: Cross-client decision coordination

---

Day 4-5: Monitor Adoption

Goal: Ensure infrastructure is being used, not bypassed

Daily Checks

Decision Usage:

cd ~/GitHub/flypilot

# Should see 2-5 new decision files per day
ls -lt docs/planning/decisions/ | head -10

# Count decisions this week
ls docs/planning/decisions/ | wc -l

# Target by end of week: 10+ decisions

Sync Operations:

# Should see cron jobs running
tail -100 ~/logs/flypilot-sync.log | grep "Rate Limit Statistics"

# Target: Zero rate limit errors
grep "Rate limit hit" ~/logs/flypilot-sync.log

Research Notes:

# Should see findings being documented
tail -20 docs/planning/research-notes.md

# Target: 10+ entries stored
wc -l docs/planning/research-notes.md

WPEngine Alerts:

# Should see alerts running every 15 min
tail -100 ~/logs/flypilot-wpengine-alerts.log

# Target: less than 10 SSL alerts per run, no backup/disk critical alerts
grep -c "Critical" ~/logs/flypilot-wpengine-alerts.log

---

Red Flags to Watch For

❌ No decisions being recorded

• Issue: Agents bypassing workflow
• Fix: Re-emphasize mandatory nature, add to role checklists

❌ Custom API clients appearing

• Issue: Agents not using sync profiles
• Fix: Remove custom clients, redirect to sync profiles

❌ Rate limit errors

• Issue: Rate limiting not being used
• Fix: Review sync profile usage, adjust limits if needed

❌ Empty research notes

• Issue: Research not being preserved
• Fix: Emphasize documentation in research role

❌ Infrastructure bypasses in code

• Issue: Developers building custom solutions
• Fix: Code review enforcement, update contribution guidelines

---

Green Lights to Celebrate

✅ Decisions growing steadily

• Indicator: Cross-orchestrator learning happening
• Celebrate: Share interesting decisions in Slack

✅ Zero rate errors

• Indicator: Rate limiting working perfectly
• Celebrate: Highlight performance improvement

✅ Research notes being reused

• Indicator: Reduced duplicate research
• Celebrate: Show time savings from finding an answer already in notes

✅ Slack alerts working

• Indicator: Proactive monitoring operational
• Celebrate: Share first caught issue that was auto-detected

✅ Sync performance improvements

• Indicator: Incremental sync reducing API calls 90%+
• Celebrate: Calculate API cost savings

---

Week 1 Success Metrics

By end of Week 1 (2026-02-10), we should see:

Metric	Target	Measurement	Current
Decisions recorded	10+	ls docs/planning/decisions/	0
Orchestrators trained	10+	Training log	0
Sync operations	Zero errors	grep "error" ~/logs/flypilot-sync.log	-
Research notes	10+ entries	wc -l docs/planning/research-notes.md	0
WPEngine alerts	<10 per run	Alert logs	-
Infrastructure bypasses	0	Code reviews	-
Rate limit hits	0	Sync logs	✅ 0

---

Common Pitfalls to Avoid

Pitfall 1: "Too Busy to Follow Checklist"

Symptom: Agents skip decision search to "save time" Reality: Spend 2x as long reinventing the wheel Solution: The checklist SAVES time by preventing duplicate work

Pitfall 2: "Decision Recording Feels Slow"

Symptom: Agents skip recording decisions Reality: Explaining the same decision 5 times later takes longer Solution: Record once, reference forever

Pitfall 3: "Not Sure What to Record"

Symptom: Agents only record "big" decisions Reality: Small decisions matter too (e.g., "use sync profile X for task Y") Solution: Record ANY choice between 2+ alternatives

Pitfall 4: "Forgetting to Search First"

Symptom: Conflicting decisions across orchestrators Reality: Searching reveals past decisions and prevents conflicts Solution: Make it muscle memory — search before deciding

Pitfall 5: "Building Custom Solutions"

Symptom: New API clients, custom sync scripts appear Reality: Infrastructure already exists and is production-ready Solution: Tool escalation protocol — use existing before building new

---

Week 2 Preview

Goals:

• Measure impact (API calls reduced, decisions reused, time saved)
• Refine workflows based on Week 1 feedback
• Expand to remaining orchestrators (all 22 total)
• Build decision library (target: 30+ decisions)
• Add infrastructure usage dashboards to fp-interface

Success Indicators:

• Agents proactively searching decisions without being reminded
• Infrastructure usage becomes automatic, not forced
• New orchestrators ask "what decisions have we made about X?"
• Cross-orchestrator learning evident in decision patterns

---

Remember

The Infrastructure is MANDATORY

The infrastructure is MANDATORY, not optional.

The goal of Week 1 is to make it HABITUAL.

By Week 2, agents should:

• Search for decisions automatically before starting work
• Record decisions as markdown files without being reminded
• Document research findings by default
• Use sync profiles instead of custom API calls
• Refer to past decisions when making new choices

This is cultural change, not just technical deployment.

Week 1 is about building the habit. Week 2 is about measuring the impact.