9. Scheduling & Running Scripts

What & why

A backup or cleanup script only earns its keep if it runs on a schedule without you. This lesson covers schtasks (create/query/delete scheduled tasks from the command line), how to launch things with start, elevation with runas, and the environment surprises that make "works when I run it, fails at 2 AM" happen.

schtasks — scheduled tasks from batch

Create

schtasks /Create /TN "MyBackup" /TR "\"%~dp0backup.bat\" --quiet" /SC DAILY /ST 02:00 /RL HIGHEST /F

Switch	Meaning
/TN	task name (use a folder: /TN "\MyOrg\MyBackup")
/TR	the command to run — quote the script path, and escape inner quotes as \"
/SC	schedule: MINUTE HOURLY DAILY WEEKLY MONTHLY ONCE ONLOGON ONIDLE ONSTART
/ST	start time HH:MM (24h)
/MO	modifier: /SC MINUTE /MO 15 = every 15 min; /SC WEEKLY /MO 1 /D MON,WED,FRI
/RL HIGHEST	run elevated (needs admin to create)
/RU /RP	run-as user / password (/RU SYSTEM for the system account, no password)
/F	force: overwrite an existing task with the same name
/IT	interactive — only runs when that user is logged on

Query

schtasks /Query /TN "MyBackup" /V /FO LIST
schtasks /Query /FO CSV /NH | findstr /i "backup"

Run now (test it)

schtasks /Run /TN "MyBackup"
rem then check the last result:
schtasks /Query /TN "MyBackup" /V /FO LIST | findstr /i "Last Result"

"Last Result" 0 = success; 267009 = currently running; 267011 = never run; other codes map to your script's exit /b.

Delete

schtasks /Delete /TN "MyBackup" /F

Self-installing script pattern

@echo off
if /i "%~1"=="--install" (
    schtasks /Create /TN "\MyOrg\Backup" /TR "\"%~f0\" --run" /SC DAILY /ST 02:00 /RL HIGHEST /F
    exit /b %errorlevel%
)
if /i "%~1"=="--uninstall" (
    schtasks /Delete /TN "\MyOrg\Backup" /F
    exit /b %errorlevel%
)
if /i "%~1"=="--run" goto :run
echo Usage: %~nx0 [--install ^| --uninstall ^| --run]
exit /b 1

:run
rem ... the actual work ...

The "works for me, fails scheduled" checklist

A scheduled task runs in a very different context than your interactive session:

1. Current directory is C:\Windows\System32, not the script's folder. → Use %~dp0 for every path, or pushd "%~dp0" at the top.
2. No mapped network drives. Z:\ doesn't exist for the SYSTEM account or a non-logged-on user. → Use full UNC paths (\\server\share\...), or net use / pushd \\server\share inside the script.
3. A leaner PATH and different env vars. Don't assume git, python, node are on PATH — call them by full path or set PATH explicitly.
4. No console / no interactivity. pause, timeout (without /nobreak... actually even then it can error), set /p, choice may hang or fail. Gate them behind an --interactive flag.
5. %TEMP% / %APPDATA% point elsewhere (SYSTEM profile). Write logs to an explicit path.
6. Different user = different permissions. SYSTEM can't reach a user's OneDrive; a service account may lack "log on as batch job."
7. UAC / elevation. A task with /RL HIGHEST runs elevated; your interactive test may not have been (or vice versa).

Defensive header

@echo off
setlocal EnableExtensions
pushd "%~dp0" || (echo cannot cd to script dir & exit /b 1)
set "LOG=%~dp0logs\run_%RANDOM%.log"
if not exist "%~dp0logs" md "%~dp0logs"
call :main >> "%LOG%" 2>&1
set "rc=%errorlevel%"
popd
exit /b %rc%

start — launch and (optionally) don't wait

start "" "notepad.exe"                    &rem launch, don't wait (note the empty "" title!)
start "" /wait "installer.exe" /silent     &rem wait for it to finish
start "Build" /min cmd /c "build.bat"       &rem minimized, titled window
start "" /b longtask.exe                    &rem no new window (background in same console)
start "" "https://example.com"              &rem open a URL / document with its default app

The empty \"\" first argument

start treats the first quoted argument as the window title. start "C:\My App\app.exe" tries to run a window titled C:\My App\app.exe and fails. Always: start "" "C:\My App\app.exe".

start /wait is how you run installers sequentially and check each result:

start "" /wait msiexec /i app.msi /qn
if errorlevel 1 ( echo install failed & exit /b 1 )

Elevation

Batch can't elevate itself directly. Options:

Detect whether you're elevated

net session >nul 2>&1
if errorlevel 1 (
    echo This script needs to run as Administrator.
    exit /b 1
)

(fltmc or openfiles also work as admin probes.)

Self-elevate via PowerShell

net session >nul 2>&1 || (
    powershell -NoProfile -Command "Start-Process -Verb RunAs -FilePath '%~f0' -ArgumentList '%*'"
    exit /b
)
echo Now running elevated.

This pops a UAC prompt and relaunches the script elevated in a new window.

runas (different user, prompts for password, no /savecred for admin)

runas /user:DOMAIN\admin "cmd /c \"%~f0\" %*"

runas can't pass a password on the command line (by design) and can't be used non-interactively for elevation — that's what schtasks /RU or a service is for.

cmd /c vs cmd /k

cmd /c script.bat        &rem run it, then the new cmd exits
cmd /k script.bat        &rem run it, then LEAVE the window open (great for debugging — see Advanced 9)

Common mistakes

• Relative paths in a scheduled script — cwd is System32. Use %~dp0 / pushd "%~dp0".
• Relying on mapped drives in a task — use UNC.
• start "C:\path with space\app.exe" — the path becomes the title. Add start "" ....
• schtasks /TR without quoting/escaping — path with spaces breaks; inner quotes need \".
• pause/set /p in a scheduled run — hangs the task until it times out.
• Assuming git/python on PATH in the task's environment.
• Not capturing output — a scheduled failure with no log is un-debuggable. Redirect >> log 2>&1.
• /RL HIGHEST but the run-as account isn't admin — task runs un-elevated silently.
• Testing elevated, deploying un-elevated (or vice versa) — mismatched permissions.

Exercises

1. Make hello.bat write %date% %time% ran as %username% in %cd% to a fixed log path. Schedule it every 5 minutes with schtasks /SC MINUTE /MO 5. Inspect the log — note the %cd%.
2. Add --install / --uninstall / --run handling to a script so it can register and remove its own scheduled task.
3. Write an "am I elevated?" check and have the script self-elevate via the PowerShell RunAs trick if not.
4. Use start "" /wait to run two commands in sequence and stop if the first fails.
5. Query your new task's "Last Result" and translate 0 / non-zero into a friendly message.
6. Break it on purpose: write a script that uses a relative path .\data\in.txt, schedule it, and confirm it fails; then fix it with %~dp0.

Recap

• schtasks /Create /TN ... /TR "\"%~f0\" args" /SC DAILY /ST 02:00 /RL HIGHEST /F — quote the path, escape inner quotes.
• /Query /V /FO LIST, /Run, /Delete /F round out the lifecycle; "Last Result" is your script's exit code.
• Scheduled context ≠ interactive: cwd is System32, no mapped drives, lean PATH, no console, different profile. Use %~dp0, UNC paths, full tool paths, and log everything.
• start "" ... (empty title!), /wait to block, /b for no window.
• Batch can't self-elevate; detect with net session, relaunch via PowerShell Start-Process -Verb RunAs.

Next: Capstone: backup utility →