Skip to content

atlas-portfolio :: OneDrive for Business backup playbook

Set up: 2026-05-25 by Atlas at Sush's request. Why this exists: Sush asked "do we have a backup for this repo in OneDrive for Business that is my corp OneDrive — if not after completing the whole polish work can you please add a backup of this repo to OneDrive for business pls to my Microsoft OneDrive and document everything in learn-doc."

TL;DR
Source C:\ssClawy\atlas-portfolio\
Destination (corp OneDrive) C:\Users\ssutheesh\OneDrive - Microsoft\_AtlasPortfolio-backup\
Size ~28 MB (data/, dist/, src/, scripts/, styles/, docs/, top-level files)
Excluded node_modules/, .git/, .copilot/, *.log
Mode Robocopy /MIR (mirror — deletes at dest what's gone at source)
Schedule Sundays 06:30 NZST (30 min before AtlasPortfolioFullRefresh at 07:00)
Task name \Atlas\AtlasPortfolioOneDriveBackup
Script C:\ssClawy\atlas-portfolio\scripts\backup-to-onedrive.ps1
Manual run pwsh C:\ssClawy\atlas-portfolio\scripts\backup-to-onedrive.ps1
Dry run pwsh C:\ssClawy\atlas-portfolio\scripts\backup-to-onedrive.ps1 -DryRun

Why corp OneDrive (not personal OneDrive, not GitHub alone)

Backup tier Status Reason
GitHub (ssutheesh_microsoft/atlas-portfolio) ✅ Active All source code, BUILD-LOG, scripts. DOES NOT include data/ or dist/ — gitignored per Standing Rule #4 (MS Confidential data stays on local disk).
Corp OneDrive (OneDrive - Microsoft) ✅ Active (this playbook) Full working dir INCLUDING data/ (snapshot-customers-*.json, lynx-snapshots/, etc.) and dist/index.html + per-account HTML. This is the only backup of the MICROSOFT CONFIDENTIAL data layer.
~~Personal OneDrive (OneDrive / "personal Microsoft account")~~ ⚠️ Pre-existing — needs cleanup A backup exists at C:\Users\ssutheesh\OneDrive\CopilotCLI_Backups\ssclawy-mirror\atlas-portfolio\. This is the wrong location for MS Confidential data per Standing Rule #4. Recommend Sush deletes it OR moves it to the corp OneDrive location. See "Cleanup TODO" below.

What's in the backup

_AtlasPortfolio-backup/
├── BUILD-LOG.md               (~230 KB · phase-by-phase history)
├── README.md                  (~7 KB)
├── .gitignore                 (~0.5 KB)
├── package.json + package-lock.json
├── data/                      (~12 MB · MICROSOFT CONFIDENTIAL)
│   ├── snapshot-customers-YYYY-MM-DD.json   (MSX customer snapshots)
│   ├── lynx-snapshots/                       (per-tenant Lynx scrapes)
│   ├── atlas-insights-cache.json             (LLM-generated commentary cache)
│   └── ...
├── dist/                      (~14 MB · rendered HTML, contains real customer data)
│   ├── index.html             (main dashboard)
│   └── account/<tpid>.html    (54 per-customer pages)
├── src/                       (~0.4 MB · source JS modules)
├── styles/                    (~0.1 MB · CSS)
├── scripts/                   (~0.3 MB · build / refresh / smoke / probes)
├── docs/                      (~0.07 MB · misc project docs)
└── reference/                 (~0.3 MB · PAC mockups for design reference)

What's NOT in the backup (and why)

Excluded Reason
node_modules/ (~15 MB) Reproducible from package.json via npm install.
.git/ The git repo IS the source-code backup (lives on GitHub).
.copilot/ Session/local state, not project content.
*.log Transient.

How to verify the backup is working

  1. Check OneDrive sync status — taskbar tray icon should show "Up to date" (cloud icon).
  2. Inspect the dest folder directly: 
    Get-ChildItem "C:\Users\ssutheesh\OneDrive - Microsoft\_AtlasPortfolio-backup\" |
  Select-Object Name, LastWriteTime
  3. Check the scheduled task's last run: 
    Get-ScheduledTaskInfo -TaskName "AtlasPortfolioOneDriveBackup" -TaskPath "\Atlas\" |
  Select-Object LastRunTime, LastTaskResult, NextRunTime
    LastTaskResult: 0 = success.
  4. Check the script log: 
    notepad "$env:TEMP\atlas-portfolio-backup-$(Get-Date -Format 'yyyy-MM-dd').log"
  5. Verify on OneDrive web — sign in to office.com → OneDrive → _AtlasPortfolio-backup folder should be browsable.

How to restore from the backup

If the local working directory at C:\ssClawy\atlas-portfolio\ is lost or corrupted:

# 1. Clone source code from GitHub (excluding data/dist which are gitignored):
gh repo clone ssutheesh_microsoft/atlas-portfolio C:\ssClawy\atlas-portfolio

# 2. Restore data/, dist/, and other non-tracked files from OneDrive backup:
robocopy "C:\Users\ssutheesh\OneDrive - Microsoft\_AtlasPortfolio-backup" `
         "C:\ssClawy\atlas-portfolio" `
         /E /XD .git /R:1 /W:1

# 3. Re-install npm deps:
cd C:\ssClawy\atlas-portfolio
npm install

# 4. Verify:
node scripts\build.mjs
node scripts\refresh-qa.mjs   # must be 25/25 green

Standing rule honoured (Microsoft Confidential data)

Per atlas-portfolio Standing Rule #4 (BUILD-LOG): "Microsoft Confidential data stays on local disk. data/ and dist/*.html are gitignored."

The corp OneDrive for Business is an acceptable storage location for MS Confidential data — it's tenant-managed, ACL-controlled to Sush only, complies with Microsoft data-residency policy. Personal OneDrive is NOT — it's a consumer service outside the corp tenant.

Cleanup TODO (Sush to action)

A pre-existing backup was found in PERSONAL OneDrive:

C:\Users\ssutheesh\OneDrive\CopilotCLI_Backups\ssclawy-mirror\atlas-portfolio\

This contains MICROSOFT CONFIDENTIAL data in the wrong location. Recommendation: delete it. The corp OneDrive backup (this playbook) supersedes it.

# Inspect first:
Get-ChildItem "C:\Users\ssutheesh\OneDrive\CopilotCLI_Backups\ssclawy-mirror\atlas-portfolio" | Select-Object Name, Length

# When ready to delete:
Remove-Item -Recurse -Force "C:\Users\ssutheesh\OneDrive\CopilotCLI_Backups\ssclawy-mirror\atlas-portfolio"
# AND: remove the whole ssclawy-mirror tree if it ONLY contains atlas-portfolio

Implementation details for future Atlas

  • Why robocopy and not Copy-Item? Robocopy handles long paths, retries, multi-thread, and produces a clean exit-code contract (0=no change, 1=copied, 2=extras-mirrored, ≥8=failure). Copy-Item would either fail or silently skip on long-path edge cases.
  • Why /MIR (mirror)? Re-runs delete anything at dest that's gone from source. This means snapshots from deleted lynx scrapes etc. don't accumulate forever in the backup. If Sush wants point-in-time snapshots, that's a different mechanism (would do dated folders like _AtlasPortfolio-snapshots\2026-05-25\).
  • Why Sunday 06:30 not 07:00? The existing AtlasPortfolioFullRefresh task fires Sunday 07:00 NZST. Running the backup 30 min before snapshots the state BEFORE the refresh starts. This way:
  • If the refresh fails, the backup still has the previous-week good state.
  • If the refresh succeeds, the next week's backup at 06:30 will pick up the fresh data.
  • Why exclude node_modules? Reproducible from package.json. Saves ~15 MB per run and OneDrive sync cycles.
  • Why no *.log excluded but the script writes a log to $env:TEMP? Logs are in %TEMP%, outside the repo — already excluded.
  • OneDrive auto-sync — the corp OneDrive client uploads changes to the cloud automatically. No need for a separate "push to cloud" step. The robocopy is the local-disk→OneDrive-folder step; OneDrive handles cloud sync.

Audit trail

Date Event
2026-05-25 Initial backup created. ~28 MB / 247 files mirrored. Scheduled task registered (\Atlas\AtlasPortfolioOneDriveBackup, Sun 06:30 NZST). Script at scripts/backup-to-onedrive.ps1. Standing-rule cleanup TODO logged for personal-OneDrive duplicate.

Read this whenever: - atlas-portfolio backup behaviour changes (schedule, source, dest) - Cleaning up the personal-OneDrive duplicate (see Cleanup TODO above) - Restoring atlas-portfolio after a local-disk failure (see "How to restore" above) - Auditing what's in the backup vs what should be (e.g., during a quarterly memory audit)