Durable Taskflow Orchestrator

The problem

Long-running workflows often fail when a session is interrupted or a service flakiness causes a restart from step zero. Coordinating multi-step jobs that require human approval or external webhooks usually leads to fragmented scripts and lost state.

What it does

• Persists execution state to a local ledger to enable resuming jobs from the exact point of failure.
• Orchestrates detached, multi-step processes that continue running after the terminal session closes.
• Manages automatic retries with exponential backoff and jitter for flaky network actions.
• Handles asynchronous pauses for webhooks or human approvals with configurable timeouts.
• Generates detailed audit trails documenting every transition, timestamp, and attempt count.

Frameworks & tools

Python, taskflow-cli, YAML for state machine definitions, and standard CLI tools like curl and mail.

Why this beats prompting it yourself

Prompting an AI to manage state often results in fragile logic that cannot handle real-world interruptions. This skill provides a structured, command-line driven protocol for durability, ensuring your agent follows a verifiable state machine rather than guessing the next step.

Use cases

• Automating multi-day billing cycles that require manual invoice approval.
• Data scraping pipelines that need to survive network drops and process restarts.
• CI/CD style deployments where later steps depend on external webhook signals.
• Complex system migrations that require a verifiable audit trail of every completed action.

Known limitations

Not suitable for short synchronous tasks, sub-second timing requirements, or heavy data movement between steps. Requires local state directory management for durability.