[Bug] Scheduled routine consumes entire token quota in single execution without safeguards
Bug Description
Subject: Scheduled routine (Claude Code Remote) consumed my entire token quota in a single run
Body:
Hello,
I'm reporting unexpectedly high token consumption from a single Claude Code Remote scheduled trigger run, which exhausted my token quota in one execution.
What happened:
I manually triggered a scheduled routine to validate its configuration (a dry run). The routine's single agent session was configured to clone 32 repositories (listed as sources) and run three review tasks across all of them. This one run consumed the large majority of my token quota — far beyond what I anticipated for a validation run.
Traceable details:
- Session ID: cse_01C8eDo1SKMk56GNiF7jLaP1
- Trigger ID: trig_01DgFKgzzASAmr2kGaWRrnk8 (name: veille-hebdo)
- Triggered manually on 2026-07-11, ~08:09 UTC
- Likely cause: a single-session design cloning 32 repositories and processing 3 review tasks over all of them, with no context isolation between tasks.
My questions:
- Can you confirm the token usage attributed to this specific session, and whether it aligns with the work actually performed?
- Given this was a one-off validation run whose cost was not foreseeable from the configuration, would you consider a credit/adjustment for this session's consumption?
- Is there any built-in safeguard (per-run token cap or budget limit) for scheduled Claude Code Remote triggers that I can enable to prevent a single run from draining my quota again?
I've since disabled the trigger to prevent any further runs. Thank you for your help.
Best regards,
Sébastien
Environment Info
- Platform: darwin
- Terminal: ghostty
- Version: 2.1.197
- Feedback ID: 8e99d083-a441-48aa-be71-61b082d2bdee
Errors
[]