# Daily Team Audit Protocol
*Created: 2026-03-30 | Evolving document — update as needed*

## The Problem This Solves
Rivet runs a nightly self-audit. But the fleet needs a **daily health check** that catches problems before they compound. Tonight proved it: Builder stalled for 28 hours, Herald ignored his inbox, Harper chased a dead grant for 30 days. None of that should have lasted more than 24 hours.

## The 10-Minute Morning Protocol
Every morning (after Michael wakes, before his window), Rivet runs this:

### 1. Output Quality Check (3 min)
- Rotate through one agent per day for deep review
- Judge: Is their output actually useful, or just "alive"?
- Fix or escalate. No status reports to Michael.

### 2. Blocker Scan (3 min)
- Who is stalled and why?
- Fix within 24h or escalate to Michael with a decision ready

### 3. SSOT Drift Check (2 min)
- Does the vault match reality?
- If Builder built something locally, did it sync?
- If Michael made a decision, did it propagate?
- Fix drift same day

### 4. Escalation Filter (2 min)
- Only bring Michael the 1-2 things that genuinely need his decision
- Everything else: handle it
- Zero escalation culture by default

## The Goal
- 84 alerts → 0
- 977 decisions pending → 2 per day max
- Builder stalls → caught within 4 hours, not 28
- Dead ends (grants) → killed within 48h, not 30 days

## Evolution Log
- **2026-03-30:** Created. Michael authorized evolution without fear.