Gabriel Koerich Orch

Define recurring work with cron expressions. Jobs create tasks on schedule (or run bash commands directly).

Commands

orch job list                                                # list scheduled jobs
orch job add "0 9 * * *" "Daily report" -b "what to do"      # task job (positional: schedule, title)
orch job add "@hourly" "Ping" -t bash -c "./scripts/ping.sh" # bash job
orch job enable <id>
orch job disable <id>
orch job remove <id>
orch job tick                                                # run one scheduler tick
orch job run <id> [-p repo-slug]                             # run a job immediately, ignoring schedule

Job Types

How It Works

1. Jobs are loaded from two sources:
   • Per-project .orch.yml under the jobs: key
   • Per-project prompts/jobs/*.md files with frontmatter (id, schedule, title, type, enabled, external, agent, command, dir)
2. The engine's scheduler checks cron schedules on each tick (tick interval configurable via engine.tick_interval).
3. When a schedule matches, a task is created (or command is run)
4. Jobs skip if a previous task from the same job is still in-flight (active_task_id)
5. Job-created tasks get scheduled and job:{id} labels

Cron Expressions

Standard cron syntax with aliases:

┌───────── minute (0-59)
│ ┌─────── hour (0-23)
│ │ ┌───── day of month (1-31)
│ │ │ ┌─── month (1-12)
│ │ │ │ ┌─ day of week (0-6, Sun=0)
│ │ │ │ │
* * * * *

Aliases: @hourly, @daily, @weekly, @monthly, @yearly

Example

# Run code quality check every weekday morning
orch job add "0 9 * * 1-5" "Code quality review" -b "Run lints, check for TODO items, update docs"

# Database backup every night (bash, no agent)
orch job add "0 2 * * *" "DB backup" -t bash -c "pg_dump mydb > /backups/nightly.sql"

Jobs integrate with GitHub sync — job-created tasks get synced to GitHub issues like any other task.