Automate with Loops, Watch, and Schedules

Flux provides three session-based automation mechanisms. Flux must remain open for all three.

#Scenario: monitor tracker work

#Repeat an agent check

While investigating generated data, ask Flux to recheck it every 15 minutes:

/loop 15m inspect data/deprecations.json and report shutdowns in the next 60 days

Stop it with /loop stop.

#React to scraper edits

The repository already has checks. Run the command once yourself, then watch the scraper directory:

/watch scraper python -m pytest -q

Stop it with /watch stop. Watch commands run directly in your shell, so use only commands you trust.

#Save a timed review

Create a weekday morning prompt:

/schedule add "0 9 * * 1-5" review data/deprecations.json and summarize shutdowns in the next 30 days

Use /schedule list, then enable, disable, or remove a task by ID.

#Choose the right mechanism

Need	Use
Repeat an agent prompt during this session	/loop
Run a command after a file changes	/watch
Save a cron prompt that runs while Flux is open	/schedule
Run after logout or in CI	An external scheduler with flux run or flux -P

#Expected result

Tracker reviews repeat at the right trigger without treating Flux as an unattended system service.

Next: Use Memory Across Sessions.