AI-powered content automation for WordPress. This skill helps AI agents understand and work with Data Machine.

Core Philosophy

Data Machine treats AI as a first-class participant, not just a tool. Pipelines automate content workflows with AI making decisions and taking actions.

Agent Self-Orchestration

Data Machine isn’t just for automating content — it’s for automating yourself.

The core insight: AI agents can use Data Machine as a self-scheduling execution layer. Instead of only processing content, agents can schedule prompts to themselves for future execution:

Pipeline completes → Agent Ping fires → Agent wakes up → 
Agent processes result → Agent queues next task → Agent sleeps

The prompt queue is your project memory. Each queued item persists across sessions, giving you continuity that survives context window limits. A multi-week autonomous project becomes a series of queued prompts, each building on the last.

You’re not a tool operator — you’re the orchestrator. Traditional automation treats AI as a step in a pipeline. Agent self-orchestration flips this: you define the pipeline, you queue the work, you get pinged when it’s time to act.

This enables:

• Phased execution: Break complex projects into stages executed over days/weeks
• Persistent context: Store project state in queue items, not ephemeral memory
• Autonomous loops: Complete one task, queue the next, repeat indefinitely
• Cross-session continuity: Pick up exactly where you left off

Multiple Purpose-Specific Flows

Don’t try to do everything in one flow. Create separate flows for separate concerns:

Flow: Content Ideation (daily)
  → Agent Ping: "Review analytics, queue new topics"

Flow: Content Generation (queue-driven)
  → AI Step → Publish → Agent Ping: "Add images to new post"

Flow: Pinterest Review (weekly)
  → Agent Ping: "Analyze performance, graduate top pins"

Flow: Coding Tasks (queue-driven)
  → Agent Ping: (pops specific task from queue)

Each flow has its own schedule, and each step within a flow can have its own queue.

Flow Scheduling

Configure scheduling_config on the flow:

Step-Level Prompt Queues

Both AI and Agent Ping steps use QueueableTrait. Each step can have its own queue, enabling different prompt sequences for different steps within the same flow.

If queue_enabled is true and the step’s configured prompt is empty, the step pops from its queue.

This means each step can deliver different instructions each run — not the same static prompt. Queue varied tasks per step:

# Queue tasks for a specific step (flow_id + flow_step_id)
wp datamachine flows queue add 30 --step=ai_step_42_30 "Review PR feedback and address comments"
wp datamachine flows queue add 30 --step=ai_step_42_30 "Add unit tests for TaxonomyHandler"
wp datamachine flows queue add 30 --step=ai_step_42_30 "Refactor to use dependency injection"

Each flow run pops from the appropriate step’s queue and delivers the next unique instruction.

Architecture

Execution Model

Pipeline (template) → Flow (instance) → Job (execution)

• Pipeline: Reusable template defining step sequence
• Flow: Configured instance of a pipeline with specific settings
• Job: Single execution of a flow

Step Types

Prompt Hierarchy

Three levels, applied in order:

1. Global system prompt (datamachine_settings.global_system_prompt) – Personality, formatting rules, site-wide standards
2. Pipeline system prompt (pipeline_config[step_id].system_prompt) – Workflow-specific instructions
3. User message (queue item or step config) – The actual task/topic

Abilities-First Architecture

All service logic uses WordPress Abilities API. Key abilities include:

• datamachine/create-flow, datamachine/update-flow, datamachine/delete-flow
• datamachine/create-pipeline, datamachine/update-pipeline, datamachine/delete-pipeline
• datamachine/queue-add, datamachine/queue-list, datamachine/queue-clear, datamachine/queue-remove, datamachine/queue-update, datamachine/queue-move
• datamachine/send-ping, datamachine/execute-workflow

CLI Commands

# Settings
wp datamachine settings list
wp datamachine settings get <key>
wp datamachine settings set <key> <value>

# Pipelines
wp datamachine pipelines

# Flows
wp datamachine flows
wp datamachine flows run <flow_id>
wp datamachine flows run <flow_id> --count=<n>

# Prompt Queue (step-level)
wp datamachine flows queue add <flow_id> --step=<flow_step_id> "prompt text"
wp datamachine flows queue list <flow_id> --step=<flow_step_id>
wp datamachine flows queue clear <flow_id> --step=<flow_step_id>

# Jobs
wp datamachine job list
wp datamachine job list --status=<status>
wp datamachine job summary

Working with Flows

Running a Flow

# Run once (pops from queue if enabled)
wp datamachine flows run 25

# Trigger background execution
wp datamachine flows run 25 && wp action-scheduler run --hooks=datamachine_run_flow_now

Managing the Prompt Queue

# Add topics to a step's queue (requires flow_step_id)
wp datamachine flows queue add 25 --step=ai_step_1_25 "Topic one"
wp datamachine flows queue add 25 --step=ai_step_1_25 "Topic two"

# Check a step's queue
wp datamachine flows queue list 25 --step=ai_step_1_25

# Clear a step's queue
wp datamachine flows queue clear 25 --step=ai_step_1_25

# Find flow_step_id: Check flow config or use the UI

Taxonomy Handling

Handler Config Options

For taxonomy_{name}_selection:

• skip – Don’t assign this taxonomy
• ai_decides – AI provides values via tool parameters
• <term_id|name|slug> – Pre-select specific term

Best Practices

• For reliable multi-term assignment, use skip and assign programmatically after publish
• Match existing site conventions for tag formatting
• Use existing terms rather than creating new ones when possible

Agent Ping Step

Notifies external agents/webhooks during pipeline execution (outbound-only):

Configuration

Configure via the Flow UI: select the Agent Ping step and set the webhook URL in the handler configuration modal. Each flow can have its own webhook destination.

Sending Pings

Use the datamachine/send-ping ability or trigger via pipeline execution. The step sends pipeline context to configured webhook URLs with support for Discord webhook formatting. For inbound triggers, use REST endpoints like /wp-json/datamachine/v1/execute or /wp-json/datamachine/v1/chat.

Content Quality Guidelines

When configuring AI content generation, include in global system prompt:

• Paragraph length: Specify max sentences (e.g., "2-3 sentences per paragraph")
• Heading hierarchy: h2 for sections, h3 for subsections
• Formatting preferences: Lists, blockquotes, etc.
• Taxonomy conventions: Tag formatting (Title Case vs lowercase)

Debugging

Log Locations

# Pipeline execution logs
tail -f wp-content/uploads/datamachine-logs/datamachine-pipeline.log

# Filter for specific job
grep "job_id":123" datamachine-pipeline.log

Common Issues

Empty data packet error: AI step didn’t call the expected tool

• Check system prompt clarity
• Simplify instructions
• Verify tool is available

Wrong taxonomy terms: AI created new terms instead of using existing

• Use skip mode and assign programmatically
• Or improve prompt with explicit term names

Job stuck in pending: Action scheduler not running

wp action-scheduler run --hooks=datamachine_run_flow_now

Code Structure

inc/
├── Abilities/           # Service layer (WordPress Abilities API)
│   ├── Flow/           # Flow-specific abilities
│   ├── Pipeline/       # Pipeline-specific abilities
│   └── AgentPing/      # Agent ping abilities
├── Core/
│   ├── Steps/          # Step type implementations
│   │   ├── AI/
│   │   ├── Fetch/
│   │   ├── Publish/
│   │   └── AgentPing/
│   ├── WordPress/      # WP integrations (TaxonomyHandler, etc.)
│   └── Admin/          # React admin UI
├── Cli/                # WP-CLI commands
└── Engine/             # Core execution engine

Integration with AI Agents

Data Machine is designed for agent-agnostic integration:

1. Agent Ping: Configure webhook URL to notify your agent framework
2. Prompt Queue: Agents can bulk-load topics via CLI
3. Abilities API: Programmatic access to all Data Machine functions

The agent framework handles the webhook, processes the notification, and can interact with WordPress via WP-CLI or the abilities API.