Workflow Clinic

Thousands of bioinformatics workflows registered on public platforms like Dockstore and WorkflowHub cannot run on modern cloud platforms. Not because the science is wrong, but because the engineering hasn't caught up. Missing containers, hardcoded paths, HPC-specific executors, and unstructured inputs and outputs make these workflows incompatible with GA4GH's WES standard and cloud-based research platforms. Workflow Clinic fixes this automatically. It is a Python-based CLI tool with two core components. The Workflow Critic scans a workflow repository, runs a deterministic rule engine followed by an optional AI agent team, and produces a structured diagnosis.json report identifying every cloudification gap with severity levels and remediation suggestions. The Workflow Doctor reads this report, applies fixes directly to the workflow files using a 3-layer strategy (AST-based, regex-based, and AI-assisted as a last resort), and opens Pull Requests to propose the changes. GitHub interaction is purely optional examine and fix work fully offline without any credentials. When a GitHub PAT is provided, the tool can additionally publish findings as GitHub issues and open Pull Requests. Without a PAT, the same outputs are generated locally as issue.md and fix.md files.