scheduled-task-watchdoglisted
Install: claude install-skill victoriacarbonell/claude-skills
# Scheduled-task watchdog
Problem: a scheduled task that fires while the user is out of usage fails, and nothing runs it again. The user only notices days later, if at all.
Solution: one extra, lightweight scheduled task (the watchdog) that runs a few times a day, finds recurring tasks whose last run failed because of the usage limit, and fires them again.
## 1. Check what is available
You need the scheduling tools: one that lists scheduled tasks with their last run (e.g. `list_triggers`), one that creates them (e.g. `create_trigger`), one that updates them (e.g. `update_trigger`) and one that fires them on demand (e.g. `fire_trigger`). Load them with ToolSearch if they are deferred. If they are not available, tell the user in one line that this environment cannot schedule tasks and stop.
List the user's scheduled tasks first. If a watchdog already exists, update it instead of creating a second one.
## 2. Ask only what you need
Ask in a single message, with sensible defaults the user can accept as is:
- **Time zone**: use it if you already know it.
- **Check times**: default to 4 checks a day spread about 4 hours apart, starting 2-3 hours after the user's latest scheduled task (usage limits usually reset within a few hours). Use an off-the-hour minute (e.g. :47).
- **Model**: default to the cheapest available model (e.g. Haiku). The job is mechanical, so a cheaper model does it just as well and spends less of the budget.
## 3. Create the watchdog
Create a recurring s