ℹ️ Context: This plan was captured on March 5, 2026 and is ready to execute. It covers five interconnected initiatives: project inventory, workspace redesign, cross-machine strategy, a second coding agent, and Chief's email identity. A new section on multi-project Telegram channels has been added based on March 8 discussion.

SECTION A — What We're Solving

The Problems

Three pain points driving this session:

Files scattered — workspace root has orphaned docs, duplicate folders, unclear homes for reports/blog/personal content
One machine, no rules — repos live only on MacBook, chief-server has no code, no clear mental model of what lives where
One agent, all contexts — Chief handles personal life, house build, coding projects, and business development all in one conversation thread. Context bleeds between sessions and there's no project isolation

SECTION B — Five Work Streams

1 Project Inventory

First task in the session: take stock of everything that exists across MacBook repos and GitHub, then produce a clean registry.

Project	Location	Status
Provider Search (MyDocList)	MacBook ~/Repo/provider-search + Hetzner	Live
CMS Data API	MacBook ~/Repo/cms-data + Hetzner	Live
Social Media Exchange	~/workspace/social-media-exchange/	Docs ready
EHI Ignite Challenge	~/workspace/ehi-ignite-challenge/	Research phase
Fraud Detector	~/workspace/projects/fraud-detector/	Blocked (SSH)
Business Analyst Agent	~/workspace/business-analyst-agent/	Planned
House Build (Orinda)	~/workspace/house-build/	Active
Wedding	~/Repo/chief/wedding/ (MacBook)	Active
Personal Website	MacBook ~/Repo/ + workspace/personal-website/	Split — needs cleanup

⚠️ Pre-flight check needed: MacBook must be awake to run: ssh blake@100.103.7.78 "ls ~/Repo/ && ls ~/Repo/chief/" — and gh CLI is not yet authenticated on chief-server.

2 Workspace Structure Redesign

Current workspace has grown organically. Proposed reorganization around purpose:

~/.openclaw/workspace/
├── SOUL.md, USER.md, AGENTS.md, MEMORY.md   ← Chief's identity (stays at root)
├── WORK.md, PERSONAL.md, TOOLS.md           ← Context files (stays at root)
│
├── projects/                                 ← Active technical projects
│   ├── fraud-detector/
│   ├── autoresearch/
│   └── [future code projects]
│
├── ventures/                                 ← Business / product work
│   ├── social-media-exchange/
│   ├── ehi-ignite-challenge/
│   ├── business-analyst-agent/
│   └── business-concepts/
│
├── personal/                                 ← Life management
│   ├── house-build/
│   ├── wedding/
│   └── [other personal]
│
├── content/                                  ← Blog + website pipeline
│   ├── research/      ← raw research
│   ├── drafts/        ← in-progress posts
│   └── posts/         ← ready to publish → syncs to website repo
│
├── reports/                                  ← Cross-project HTML reports (keep as-is)
│   └── templates/
│
├── memory/                                   ← Daily notes (keep as-is)
│
└── config/                                   ← Ops/infrastructure notes
    ├── VOICE_WORKFLOW.md
    ├── hetzner-setup-plan.md
    └── [other infra docs]

Open Questions to Resolve

Do cross-project reports stay in reports/ flat, or get tagged/indexed?
Blog pipeline: where does content/ hand off to the website repo?
Should workspace be structured by purpose (as above) or by project?
Where do orphaned files go? (anthropic cover letter, old cron reports, etc.)

3 Cross-Machine Strategy

Proposed mental model:

Mac Mini (chief-server, 100.105.36.5)
  └── Always-on brain: OpenClaw workspace, memory, MEMORY.md
  └── Server-side project repos: CMS API, analytics, deploy targets
  └── Source of truth for Chief's config + memory

MacBook Pro (100.103.7.78)
  └── Development workstation: VS Code, active coding
  └── All dev repos: provider-search, cms-data, chief/, website
  └── SSHes to chief-server for server management

GitHub (blakethom8)
  └── Source of truth for all code
  └── chief-server pulls from GitHub for deploys
  └── MacBook pushes to GitHub from local dev

Scenario	Sync Method
Code changes	MacBook → GitHub → chief-server pulls for deploy
Memory / workspace docs	chief-server is truth; MacBook reads via SSH if needed
Reports / HTML files	chief-server; publish to GitHub Pages for mobile
One-off file transfers	rsync or scp via Tailscale

4 Second Agent — Coding-Focused Profile

OpenClaw natively supports multiple agents with full isolation. Each agent gets its own workspace, SOUL.md, AGENTS.md, auth profiles, and session store.

ℹ️ How to create: openclaw agents add coding — creates an isolated coding agent profile. Can bind to a different Telegram bot or different channel routing.

	Chief (main)	Coding Agent
Purpose	Personal + business life + strategy	Code, servers, deploys, repos
Context loaded	MEMORY.md, personal context, all projects	CLAUDE.md per project, server configs, CI/CD
Channel	Main Telegram bot	Separate Telegram bot or slash command
Auth	Full (Google, Telegram, GitHub)	Restricted: GitHub + exec tools
Session history	Full personal context	Project-scoped only

Design Questions

Which channel does the coding agent live on? (Separate bot token? Slash command prefix?)
How do we hand off tasks from Chief → coding agent? (Explicit "spawn" command?)
Should coding agent have full tool access or restricted to code/exec only?

SECTION C — NEW: Multi-Project Telegram Channels

✅ New idea (March 8): Use separate Telegram channels/bots for different projects — each with its own persistent session history — so you can track full context for individual work streams independently.

The Problem This Solves

Right now, every conversation — whether it's house build, fraud detection, EHI Ignite, or personal logistics — all runs through the same Chief session. That means:

Long threads lose context as projects compete for window space
Returning to a project (e.g., "what were we doing on the fraud detector?") requires memory archaeology
No clean "this is the complete history of project X" view

Proposed Architecture

Channel / Bot	Purpose	Agent	Session history
📱 Main Chief (existing)	Personal life, logistics, quick tasks	main	Full personal context
💻 Coding Agent	All dev work: provider-search, cms-api, deploys	coding	Code + server context only
🏥 Healthcare Venture	EHI Ignite, fraud detector, CMS pipeline	coding or dedicated	Healthcare project context
🏠 House Build	Orinda project tracking, wetland fight, permits	main or dedicated	Full house build history

How OpenClaw Supports This

Multiple Telegram bots: Each agent can have its own bot token → separate Telegram conversations
Same bot, different groups: Add bot to different Telegram groups; Chief routes by chat ID
Session isolation: Each chat ID = separate session history = no bleed between projects
AGENTS.md per agent: Each agent gets its own context files, workspace, and loaded projects

Simplest Path Forward

Option A (easiest, no new bots): Create separate Telegram groups, add existing Chief bot to each. Each group gets its own session history. Chief adapts behavior based on which group he's in (via AGENTS.md group-context rules).

Option B (cleaner, more isolated): New Telegram bot token per agent/context. Total isolation. Requires creating new bots via @BotFather and updating openclaw config.

Recommendation: Start with Option A for low-friction testing. Migrate to Option B when you have 3+ active project threads.

Questions to Decide

Which projects get their own channel first? (Fraud detector + EHI Ignite seem like natural candidates)
Should house build stay in main Chief or get its own channel?
One bot per agent, or one bot with multiple groups?
How do you want to "start" a project conversation — /new in the project channel?

SECTION D — Chief's Email Identity

Options (Ranked by Fit)

Option	What It Is	Best For	Status
C — Gmail Delegate	Blake's inbox, Chief reads/drafts, sends "on behalf of Blake Thomson"	EA model: triage, drafts, replies	Recommended start
B — Dedicated Account	chief@blakethomson.studio (domain is live)	Chief's own outreach identity	Add when needed
A — Send-As Alias	Alias off Blake's existing Gmail	Lightweight identity	Simplest but clunky
D — Cal.com for scheduling	Skip email, use scheduling links	Meeting coordination only	Doesn't solve inbox

SECTION E — Pre-Session Checklist

Before starting the actual restructure work, these need to happen:

☐ MacBook awake + SSH reachable: ssh blake@100.103.7.78 "ls ~/Repo/"
☐ GitHub inventory: gh repo list blakethom8 (need gh auth on chief-server first)
☐ Review current openclaw.json: cat ~/.openclaw/openclaw.json
☐ Check existing migration docs: MIGRATION-PLAN.md, OPENCLAW-MIGRATION-CHECKLIST.md, WORKSPACE-SETUP-CHECKLIST.md
☐ Decide: Option A or B for Telegram multi-channel

Root-Level Cleanup (Orphaned Files)

These files are sitting at workspace root and need a home:

File	Move To
Anthropic_Cover_Letter.md, Why_Anthropic_.md, Resume_.docx	anthropic-application/
CRON_REPORT_, FEATURE2_SUMMARY.md, FEATURE_PLAN_.md	Archive or delete (stale)
ryan_med_school_response.txt, sgv_ent_medical_schools.csv	inbox-cleanup/ or named project
browser-automation-learning.md	ventures/ or business-concepts/
hetzner-setup-plan.md	config/ (new folder)
VOICE_WORKFLOW.md	config/

Chief Environment Restructure Plan · Compiled March 8, 2026 · Source: ENVIRONMENT-RESTRUCTURE-PLAN.md + March 8 session notes · Ready to execute on command

🎯 Chief Environment Restructure Plan