Tasks

What Are Tasks?

A task is your conversation with decopilot. It’s where work happens—where you chat back and forth, decopilot uses tools, and context accumulates as you progress.

Tasks are accessible throughout deco CMS:

• Homepage - Your organization’s chat interface shows the current task
• Sidebar - Click the chat icon to open the side panel and continue your task
• Task history - Browse past tasks to review previous work

Every task has a scope (organization, project, or agent) that determines which instructions load into context. You can switch scopes anytime using the button at the bottom of the chat input—switch between organization-wide work, project-specific focus, or specialized agent scopes as needed.

Think of tasks as persistent workspaces where context builds up naturally as you work.

Task States and Monitoring

Tasks move through different states as work progresses. Understanding these states helps you manage work and know when tasks need your attention.

Task Status Overview

Every task has one of four statuses:

• In Progress - Active work happening, agent is executing or waiting for next message
• Requires Action - Paused, waiting for your input (tool approval or question response)
• Success - Work completed successfully
• Failed - Something went wrong or task timed out

When Tasks Require Attention

Tasks pause and transition to “Requires Action” status when:

Tool approval needed - Some tools require explicit user approval before execution. The task pauses until you review and approve or deny the tool call.

Question from agent - When decopilot needs clarification or decisions, it asks questions using the user_ask tool. The task waits for your response.

You can see all tasks requiring attention by clicking the tasks icon in the deco CMS sidebar. This view shows tasks grouped by status—letting you quickly identify which tasks need your input.

Task Timeouts

Tasks automatically timeout after 30 minutes of inactivity. When this happens:

• Task transitions from “In Progress” to “Failed” status
• You’ll see the timeout noted in task history
• You can restart work by creating a new task

Timeouts prevent tasks from staying in limbo indefinitely. If you’re working on something that naturally takes longer, just send a message or interact with the task periodically to keep it active.

Viewing Subtask Progress

Subtasks appear in the chat interface while they execute. You can see:

• Subtask status and progress
• Tools being used
• Agent reasoning (if enabled)

However, inputs are disabled during subtask execution—you can’t interact with a subtask directly. Subtasks run autonomously and return summaries to the parent task when complete.

What Are Subtasks?

A subtask is a separate conversation that branches off from your main task to handle focused work. The key difference: subtasks start with fresh context—they don’t inherit your accumulated conversation history.

This makes subtasks perfect for intensive work that would otherwise fill up your main context window.

How Subtasks Work

When you spawn a subtask:

1. Fresh start - Subtask begins with clean context (just scope instructions and system prompt)
2. Independent execution - Subtask is a regular task with restricted built-in tools (no subtask_run or user_ask )
3. Summary return - Parent task receives only a concise summary (200-500 tokens)
4. Full details preserved - Complete subtask conversation saved in task history

The parent task sees just the summary—keeping your main conversation lean—while full details remain accessible in the subtask history.

Key Differences

Subtasks can’t spawn additional subtasks (prevents infinite nesting) and can’t ask user questions (they run autonomously).

When to Use Subtasks

Subtasks shine when you need focused work that would clutter your main conversation.

Research and Analysis

Deep dives into data, exploring options, or investigating issues work well in subtasks.

Example: You’re planning inventory purchases and need demand analysis. Spawn a subtask to research sales trends, seasonal patterns, and competitor behavior—then get back just the recommendation, not the dozens of data points analyzed.

Complex Calculations

Number-crunching work that generates verbose output belongs in subtasks.

Example: Calculating optimal shipping routes across multiple warehouses. The subtask analyzes distances, traffic, carrier schedules—you get back just the routing plan, not the calculation details.

Quality Checks and Reviews

Audits, validation, and review work benefit from isolated context.

Example: Before transferring 500 SKUs between warehouses, spawn a subtask with an “Inventory Auditor” agent to verify stock levels. Get back a summary of issues found and transfers validated—keep the detailed audit work out of your main task.

Parallel Work

When you have independent operations that can run simultaneously, spawn multiple subtasks.

Example: Monthly reporting across departments—spawn separate subtasks for apparel, electronics, and home goods analysis. All run in parallel, summaries come back to your main task for synthesis.

Working with Agent Scopes

Subtasks can run in different agent scopes than your main task—this is how specialized agents collaborate.

Example: Your main task is in organization scope working on order fulfillment. You spawn subtasks using:

• Inventory Agent - Assesses stock availability
• Shipping Agent - Plans logistics and routing
• Customer Service Agent - Prepares support strategies

Each agent brings specialized expertise. Your main task coordinates the overall strategy while agents handle domain-specific details in isolation.

This pattern keeps specialized work contained while letting you benefit from agent expertise.

Common Patterns

Pattern 1: Parallel Execution

Spawn multiple subtasks for independent work that can run simultaneously.

When to use: Monthly reporting, multi-department analysis, independent validations

Benefit: Faster execution, isolated contexts prevent cross-contamination

Pattern 2: Sequential Delegation

Chain subtasks where each step builds on previous results.

When to use: Multi-stage workflows, progressive refinement, decision pipelines

Benefit: Clear progression, focused subtasks, results build toward comprehensive outcome

Pattern 3: Specialist Collaboration

Use different agent scopes for subtasks requiring specialized knowledge.

When to use: Cross-functional work, domain expertise needed, coordinated strategies

Benefit: Leverage specialized agents, integrate multi-domain insights

Practical Tips

Offload heavy lifting - Research, calculations, and analysis should happen in subtasks. Keep summaries in your main task, not verbose details.

Match scope to work - Use specialized agents for domain tasks, organization scope for general work. The right scope brings relevant instructions and context.

Work in parallel - Independent subtasks can run simultaneously. Don’t wait if the next subtask doesn’t depend on the previous one.

Name clearly - Give descriptive prompts when spawning: “Analyze Q4 apparel demand trends” rather than “Do analysis”. This helps when reviewing task history.

Review full details when needed - The complete subtask conversation lives in task history. If you need raw data or detailed analysis, open the subtask from history.

Watch for clutter signals - When your main task feels heavy or context-full, spawn subtasks for the next phase of work.

Context Benefits

Subtasks are powerful for managing your context window:

• Main task stays lean - Only summaries accumulate, not analysis details
• Fresh context for complexity - Subtasks start clean, ideal for deep work
• Longer productivity - Avoid premature compactification in main task
• Focused execution - Each subtask tackles one thing without baggage

By keeping intensive work in subtasks, your main task remains focused and productive across longer sessions.

---

Next steps: Learn about Context to understand how tasks manage your working memory, explore Tools to see what decopilot can do, or check out Agents for specialized scopes you can use with subtasks.