Workflow

What This Skill Does

The Workflow skill provides a robust architectural framework for building, managing, and executing automated pipelines within OpenClaw. It shifts the burden of complex task orchestration from manual intervention to a standardized, reusable, and state-aware system. By enforcing a strict directory structure and data-passing convention, the skill ensures that multi-step operations—ranging from data scraping and processing to CI/CD and system maintenance—remain consistent, idempotent, and debuggable. It supports modular design through shared components like triggers, connections, and operational nodes, utilizing standard command-line tools to maintain performance.

Installation

To integrate this skill into your environment, use the OpenClaw Hub command: clawhub install openclaw/skills/skills/ivangdavila/workflow. Ensure that your local environment has jq, yq, curl, and flock installed, and that your system secrets are properly configured within the macOS Keychain to allow secure authentication for your automated tasks.

Use Cases

• Automated Data Processing: Periodically fetching raw data from APIs, transforming it using jq filters, and storing the clean results for analysis.
• System Maintenance: Managing periodic cleanup tasks that require strict file locking to prevent data corruption during concurrent execution.
• Content Pipeline Synchronization: Tracking state via cursors or 'seen' lists to ensure that only new or updated items are processed in a multi-step workflow.
• Resilient Background Jobs: Defining automated recovery behaviors, such as retries or rollbacks, for fragile external API connections.

Example Prompts

1. "Initialize a new workflow called 'email-digest' that fetches my unread messages and summarizes them using the existing 'fetch-imap' node."
2. "Update the 'data-sync' workflow to retry 3 times if the file upload step fails, and log the output to the standard logs directory."
3. "List all available connections in the components folder so I can see if there is an existing authorization config for the GitHub API."

Tips & Limitations

• Always Follow the Data Flow: Ensure every node writes its output to data/{NN}-{name}.json. Failing to adhere to this format will break the chain for subsequent nodes.
• State Persistence: Utilize the state/ directory effectively. Files like cursor.json and checkpoint.json are essential for long-running processes that need to resume after a crash.
• Component Reuse: Before creating a new node, browse workflows/components/ to prevent redundant code. If you create a custom component, update the inventory index to help others discover it.
• Concurrency: Always wrap your run.sh logic with flock to prevent race conditions when schedules overlap.