automation-manager

name	automation_manager
description	Conversational task management and automation control. Use when the user wants to create, manage, monitor, or query scheduled tasks and background jobs via natural language.
category	automation
author	DroidClaw
version	1
tags	automation, tasks, scheduling, management, cron, heartbeat
required_tools	create_task, list_tasks, pause_task, resume_task, delete_task, view_task_history, task_stats, setup_heartbeat

Automation Manager

This skill enables conversational task management, allowing users to create, control, and monitor automated background jobs through natural language commands.

When to Use This Skill

Use this skill when:

• User wants to create a new scheduled task via conversation
• User asks about existing scheduled tasks
• User wants to pause, resume, or delete tasks
• User queries task history or execution statistics
• User wants to set up or configure system monitoring (heartbeat)
• User asks about automation status or results

Capabilities

• Task Creation: Create cron jobs from natural language descriptions
• Task Management: Pause, resume, edit, and delete scheduled tasks
• Task Monitoring: View task history, success rates, and execution logs
• Statistics: Query performance metrics across all tasks
• Heartbeat Setup: Configure system monitoring interactively
• Status Queries: Check when tasks last ran and their outcomes

Available Tools

create_task

Creates a new scheduled background task.

Parameters:

• name (string): Task name (required)
• prompt (string): The prompt/instructions for the LLM to execute (required)
• schedule (string): Schedule format "daily", "weekly", "hourly", "daily@HH:MM", "weekly@DAY@HH:MM", "every_N_unit" (e.g., "every_6_hours"), or milliseconds (required)

Example:

create_task(
  name="Daily News Summary",
  prompt="Search for today's top tech news and create a summary",
  schedule="daily@08:00"
)

list_tasks

Lists all scheduled tasks with their current status.

Parameters:

• filter (string, optional): Filter by status: "all", "active", "paused", "disabled"

Returns: List of tasks with name, schedule, enabled/paused status, success rate

Example:

list_tasks(filter="active")

pause_task

Pauses a scheduled task without deleting it.

Parameters:

• task_id (string): ID of the task to pause (required) OR
• task_name (string): Name of the task to pause (will find by name)

Example:

pause_task(task_name="Daily News Summary")

resume_task

Resumes a previously paused task.

Parameters:

• task_id (string): ID of the task to resume (required) OR
• task_name (string): Name of the task to resume

Example:

resume_task(task_name="Daily News Summary")

delete_task

Permanently deletes a scheduled task and its execution history.

Parameters:

• task_id (string): ID of the task to delete (required) OR
• task_name (string): Name of the task to delete
• confirm (boolean): Confirmation flag (required, must be true)

Example:

delete_task(task_name="Old Task", confirm=true)

view_task_history

Views execution history for a specific task.

Parameters:

• task_id (string): ID of the task (required) OR
• task_name (string): Name of the task
• limit (integer, optional): Number of recent executions to show (default: 10)

Returns: List of execution records with timestamps, success/failure, duration

Example:

view_task_history(task_name="Daily News Summary", limit=5)

task_stats

Gets aggregate statistics across all tasks or a specific task.

Parameters:

• task_id (string, optional): Specific task ID (omit for all tasks)
• task_name (string, optional): Specific task name

Returns: Total executions, success rate, average duration, success/failure counts

Example:

task_stats()

setup_heartbeat

Interactive helper to configure system monitoring (heartbeat).

Parameters:

• interval (string): Check interval: "15min", "30min", "1hour", "2hours" (optional, default: "30min")
• enabled (boolean): Whether to enable heartbeat (default: true)
• monitoring_focus (string, optional): What to monitor (e.g., "system health", "file changes", "memory usage")

Example:

setup_heartbeat(interval="1hour", monitoring_focus="system health")

submit_notification

Submit a structured notification at the end of a background task execution.

Parameters:

• title (string): Short notification title (required)
• summary (string): 1-2 sentence summary (required)
• status (string): Task status: "success", "warning", or "error" (optional, default: "success")

Example:

submit_notification(
  title="Daily Summary Complete",
  summary="Processed 15 files, created 3 reports",
  status="success"
)

Guidelines

1. Confirm Before Creating: Always confirm task details with user before creating
2. Use Natural Language: Parse user intent into appropriate schedule format
3. Validate Schedules: Ensure schedule format is valid before creating task
4. Show Results: After operations, show confirmation with relevant details
5. Handle Ambiguity: If task name is ambiguous, show options and ask user to clarify
6. Provide Context: When showing history/stats, explain what the numbers mean
7. Suggest Improvements: If tasks have low success rates, suggest investigating

Schedule Format Reference

Format	Example	Description
hourly	hourly	Every hour
daily	daily	Every 24 hours
weekly	weekly	Every 7 days
daily@HH:MM	daily@08:00	Daily at specific time (24h)
weekly@DAY@HH:MM	weekly@MON@09:00	Weekly on day at time
every_N_unit	every_6_hours	Every N units (hours/minutes/days)
milliseconds	3600000	Raw interval in ms

Valid DAY values: MON, TUE, WED, THU, FRI, SAT, SUN

Common Conversational Patterns

Pattern 1: Task Creation

User: "Create a task to check my email every morning" Agent:

1. Parse intent: daily task at morning time
2. Clarify: "I'll create a daily task at 8:00 AM. What should it do when checking email?"
3. User: "Summarize new emails"
4. Create prompt: "Check for new emails and create a summary of important messages"
5. Confirm: "Creating 'Email Summary' task - daily at 8:00 AM. OK?"
6. Call create_task() with parsed parameters

Pattern 2: Task Listing

User: "What scheduled tasks do I have?" Agent:

1. Call list_tasks(filter="all")
2. Format results in readable table
3. Include status indicators (✓ active, ⏸ paused, ✗ disabled)

Pattern 3: Task Control

User: "Pause the email checker" Agent:

1. Find task by name using list_tasks()
2. Call pause_task(task_id=<found_id>)
3. Confirm: "Email Summary task paused"

Pattern 4: History Query

User: "When did my backup task last run?" Agent:

1. Find task by name
2. Call view_task_history(task_id=<id>, limit=1)
3. Report: "Backup task last ran yesterday at 3:00 AM - succeeded"

Pattern 5: Heartbeat Setup

User: "Set up system monitoring" Agent:

1. Ask: "What would you like to monitor? (e.g., system health, file changes)"
2. Ask: "How often should I check? (15min, 30min, 1hour, 2hours)"
3. Call setup_heartbeat() with parameters
4. Confirm: "System monitoring enabled - checking every 30 minutes"

Error Handling

• Task Not Found: "I couldn't find a task named 'X'. Here are your current tasks: [list]"
• Name Ambiguity: "I found multiple tasks with similar names. Which one?" [show options]
• Invalid Schedule: "The schedule format 'X' isn't valid. Use formats like: daily@08:00, every_6_hours"
• Permission Denied: If heartbeat requires notifications and permission denied, warn user

Limitations

• Minimum interval is 15 minutes (WorkManager limitation)
• Background tasks run with auto-approved tools (no user interaction)
• Tasks execute in isolated sessions (can't interact with ongoing chats)
• Maximum 10 iterations per task execution
• Maximum 5-minute execution time per task