Tasks

Tasks

Tasks are the core building block of 0ct. A task defines what you want AI to do, when to do it, where to get data from, and where to deliver results.

Task Components

Every task consists of:

Component	Description	Required
Name	Human-readable identifier	Yes
Prompt	Instructions for the AI	Yes
Model	Which AI model to use	Yes
Schedule	When to run	Yes
Sources	Data connections	No
Skills	Reusable instructions	No
Destinations	Where to deliver results	No

Creating a Task

Via Dashboard

1. Navigate to Tasks in the sidebar
2. Click New Task
3. Fill in the task details
4. Click Create Task

Via SDK

import Oct from '0ct';

const client = new Oct({
  apiKey: process.env.OCT_API_KEY
});

const task = await client.promptly.tasks.create({
  name: 'Weekly Report',
  prompt: 'Generate a summary of this week\'s key metrics...',
  modelId: 'anthropic/claude-3-5-sonnet',
  frequency: 'weekly',
  scheduledDay: 'monday',
  scheduledTime: '09:00',
  isActive: true
});

Scheduling Options

Frequency Types

Hourly: Run every N hours

{
  frequency: 'hourly',
  hourlyInterval: 2  // Every 2 hours
}

Daily: Run once per day at a specific time

{
  frequency: 'daily',
  scheduledTime: '09:00'  // 9 AM
}

Weekly: Run on specific days

{
  frequency: 'weekly',
  scheduledDay: 'monday',
  scheduledTime: '08:00'
}

One-time: Run once at a specific date/time

{
  frequency: 'once',
  scheduledDate: '2026-02-15',
  scheduledTime: '15:00'
}

Timezone Handling

All times are in your organization’s configured timezone. Set your timezone in Settings → Organization.

Prompt Engineering

Best Practices

1. Be Specific: Clearly describe what you want
2. Provide Context: Explain why the task matters
3. Define Output Format: Specify how results should be structured
4. Set Boundaries: Clarify what to include/exclude

Example Prompts

Good Prompt:

Analyze our GitHub repository activity from the past week.

For each contributor, report:
- Number of commits
- Lines added/removed
- Key changes made

Summarize overall team velocity and highlight any concerning trends.
Format as a markdown report suitable for our weekly standup.

Weak Prompt:

Tell me about our GitHub.

Using Variables

Prompts can include dynamic variables:

Generate a {{frequency}} report for {{organization_name}}.
Today's date is {{current_date}}.

Attaching Sources

Sources give your task access to real-time data. When sources are attached, the AI can query them during execution.

// Attach sources when creating
const task = await client.promptly.tasks.create({
  name: 'Project Status',
  prompt: 'Summarize open issues and recent activity...',
  modelId: 'openai/gpt-4o',
  frequency: 'daily',
  scheduledTime: '09:00',
  sources: ['source_linear_abc123', 'source_github_def456']
});

// Or update existing task
await client.promptly.tasks.update(task.id, {
  sources: ['source_slack_ghi789']
});

Attaching Skills

Skills provide reusable instructions that enhance your prompts:

const task = await client.promptly.tasks.create({
  name: 'Customer Email',
  prompt: 'Draft a response to this customer inquiry...',
  modelId: 'anthropic/claude-3-5-sonnet',
  frequency: 'hourly',
  hourlyInterval: 1,
  skills: ['skill_writing_style', 'skill_product_knowledge']
});

Skills are prepended to your prompt, giving the AI consistent context.

Configuring Destinations

Destinations define where results are delivered:

const task = await client.promptly.tasks.create({
  name: 'Daily Digest',
  prompt: 'Compile today\'s key updates...',
  modelId: 'openai/gpt-4o',
  frequency: 'daily',
  scheduledTime: '18:00',
  destinations: ['dest_email_team', 'dest_slack_channel']
});

Managing Tasks

List All Tasks

const tasks = await client.promptly.tasks.list();

for (const task of tasks.data) {
  console.log(`${task.name}: ${task.isActive ? 'Active' : 'Paused'}`);
}

Update a Task

await client.promptly.tasks.update('task_abc123', {
  prompt: 'Updated instructions...',
  isActive: false  // Pause the task
});

Delete a Task

await client.promptly.tasks.delete('task_abc123');

Trigger Manual Run

Don’t wait for the schedule—run immediately:

const run = await client.promptly.tasks.run('task_abc123');
console.log('Run ID:', run.id);

Monitoring Runs

View Run History for a Task

const runs = await client.promptly.tasks.listRuns('task_abc123', {
  limit: 10
});

for (const run of runs) {
  console.log(`${run.id}: ${run.status}`);
  console.log(`Tokens: ${run.tokensUsed}`);
}

View All Runs

const allRuns = await client.promptly.runs.list({
  organizationId: 'org_abc123',
  limit: 50
});

Run Statuses

Status	Description
pending	Scheduled, waiting to execute
running	Currently executing
completed	Finished successfully
failed	Encountered an error

Error Handling

Tasks can fail for various reasons:

• Model API errors
• Source connection failures
• Destination delivery issues
• Prompt/response validation errors

Failed runs include error details:

const run = await client.promptly.runs.get('run_failed');

if (run.status === 'failed') {
  console.log('Error:', run.error);
}

Next Steps

• Connect Sources - Give tasks access to your data
• Create Skills - Build reusable instructions
• Configure Destinations - Set up delivery channels