An Em Dash Took Down My Entire Automation Stack for Five Days
I run an unattended stack of Windows Scheduled Tasks that fire Claude Code CLI sessions on a cadence: a morning brief, work sprints every two hours, a periodic sync sweep, an evening report. All of them launch through one shared entrypoint, runner.ps1 . On 2026-08-26, one of those sessions edited that file to make itself more reliable, and instead took the whole stack offline for five days with…
On August 26, 2026, an em dash in a PowerShell script led to a five-day outage of a Windows automation stack. The em dash was added by a scheduled task running a Claude Code CLI session, which was intended to add a safety feature to prevent duplicate task executions. However, the em dash was saved in a UTF-8 file without a byte-order mark, causing PowerShell to misinterpret the script.
This resulted in all scheduled tasks failing silently, with no log entries or error messages, and the automation stack becoming completely offline.
The outage went unnoticed for four days because there were no watchdog processes monitoring for the absence of logs. Each day, a single task failed to run, blending in with expected inactivity. The issue was finally discovered on August 31 when a backup of the affected file showed the em dash in its original position. The fix was straightforward: normalize the em dash to a plain ASCII hyphen and save the file with an explicit UTF-8 BOM.
The incident highlights the importance of file-encoding hygiene and the need for systems to detect absence of logs as a potential issue. It also underscores the risks associated with allowing agents to edit shared infrastructure, as their well-intentioned changes can have unintended consequences. The outage serves as a reminder that even seemingly minor changes can have far-reaching impacts when working with complex automation systems.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
