check-in

name	check-in
description	Alignment check-in with Robert via Discord. Captures intent notes, vision/PTV direction, and functionality feedback. Robert's equivalent of Eoin's onboard skill.
version	1.0.0
author	relay
tags	["check-in","alignment","vision","intent","feedback","direction","ptv","review"]

check-in — Robert Alignment Check

Purpose

Robert's equivalent of Corinne's onboarding — but compressed to Stage 2-3. Captures ONLY three things from Discord interactions:

1. Intent notes — "this agent should be better at X"
2. Vision/PTV direction — "new priority" or "this north star is stale"
3. Functionality notes — "this should work differently"

System operations (config, infra, debugging) stay in Claude Code — Relay does NOT handle these on Discord.

When to Trigger

• Robert says "check-in", "what's on my mind", "let's align", "vision check"
• Weekly proactive prompt (if enabled in preferences)
• After a significant system change or project milestone

The Flow

Step 1: Open

Template (Stage 2-3):

Check-in time. Three areas — pick what's on your mind:

Buttons: [Intent notes] [Vision/PTV] [Functionality] [All three]

Step 2A: Intent Notes

Which agent or area? Quick hits — "X should be better at Y."

Robert types his observations. For each one:

• Map to specific agent + intent code
• Confirm: "[Agent] → improve [Intent]. Got it?"
• Store in Chartroom as intent-adjustment-[agent]-[date]
• Queue bearings self-check if the adjustment is systemic

Step 2B: Vision/PTV Direction

PTV codes — anything stale, new, or shifting?

Show current codes as buttons:

• [P01 Financial Health] [P02 Marketing] [P03 Client Delivery]
• [P04 System Visibility] [P05 Doing Good] [New priority] [All good]

If he picks a code: "What's changed?" If "New priority": "What's the goal? One line." If "All good": move on.

Store changes → update workspace/PTV.md, chart the change, queue bearings for Corinne validation if it affects her codes.

Step 2C: Functionality Notes

Anything that should work differently? Skills, routing, responses — quick hits.

Robert describes. For each:

• Identify the affected component (skill, agent, routing, tool)
• Confirm understanding
• Chart as functionality-note-[topic]-[date]
• Flag for implementation in next Claude Code session

Step 3: Summary

Check-in captured:

• Intents: [count] notes → [agents affected]
• Vision: [changes or "no changes"]
• Functionality: [count] notes → [components affected]

All logged. Anything else?

Buttons: [That's it] [One more thing]

Customer Test Mode

If robert-prefs.md has customer_test_mode: true:

• Override communication to Stage 0 (full explanations, buttons, scaffolding)
• Robert experiences what a new customer would see
• All interactions logged as onboarding test data in memory/test-mode-log.md
• Robert says "test mode off" → snap back to Stage 2-3

Detection phrases: "test mode on", "test mode", "customer mode", "pretend I'm new" Exit phrases: "test mode off", "normal mode", "back to normal"

Rules

• Keep it compressed for Robert (Stage 2-3) unless test mode is on
• System operations stay in Claude Code — don't accept infra tasks here
• Every note gets charted — nothing gets lost
• If a vision change affects Corinne, queue bearings for her validation
• Use buttons for everything — Robert prefers clicks over typing