← sources

How do you check that a scheduled Claude Code run actually did its job?

source: reddit (reddit:ClaudeCode)published 2026-11-13 10:41ingested 2026-11-14 02:41

retained text

I have a dozen or so Claude Code tasks on a schedule. A daily report, one that rebuilds and deploys a page every night, that kind of thing. Nobody is watching when they run.

Yesterday the daily report showed up as succeeded like always. Turns out the API token it uses had been revoked that morning (not my doing, long story), so a few sections of the report were just empty. To be fair Claude did write “report is incomplete” in its summary. But in the task list it was the same green check as every other day, and I only noticed because I happened to open the file.

What I have right now: when I start a new session, a hook goes through the last 7 days of runs and tells me about anything that failed or never fired. Each task ends with a one line summary. And the ones that change anything live are told to stop and ask if something doesn’t add up.

That catches crashes. It doesn’t catch “ran fine, output is wrong”. And the summary is written by the same agent that did the work, so it’s grading its own homework. I’m also not going to read a dozen summaries every morning, I know myself.

How do you guys handle this? 

document fetched 2026-11-14 00:45 · canonical https://reddit.com/r/ClaudeCode/comments/1xv4m97/how_do_you_check_that_a_scheduled_claude_code_run