wdl-task-starter

Purpose

Run a local WDL task end to end and use it as a starter for task and workflow authoring.

When to use

• You need a verified OpenWDL-style local example.
• You want a small WDL task or workflow starter with JSON inputs.

When not to use

• You need cloud backends or Cromwell execution.
• You need advanced imports or large workflow graphs.

Inputs

• A WDL document
• A Cromwell-style JSON input object
• Optional run directory and summary path

Outputs

• A JSON run summary with the output JSON emitted by miniwdl
• The rendered greeting.txt file inside the run directory

Requirements

• slurm/envs/workflow-languages
• miniwdl
• udocker inside the same prefix for daemonless local execution

Procedure

1. Run python3 skills/reproducible-workflows/wdl-task-starter/scripts/run_wdl_hello.py --name Ada --workspace scratch/wdl/workspace --summary-out scratch/wdl/summary.json.
2. Inspect the WDL outputs JSON and the materialized greeting.txt; the wrapper configures a daemonless local miniwdl execution path for this toy task.
3. Use examples/hello.wdl and examples/hello-inputs.json as the base for new tasks or workflows.

Validation

• miniwdl run exits successfully.
• The output JSON includes greeting text and a file output.

Failure modes and fixes

• Missing runtime: verify slurm/envs/workflow-languages exists and python -m WDL --version works.
• Missing daemonless backend: verify slurm/envs/workflow-languages/bin/udocker --version works.
• Input mismatch: run miniwdl run --json to inspect the required input object shape.

Safety and limits

• This skill runs a local toy workflow only.
• Extend runtime blocks carefully before using external containers or large file inputs.

Provenance

• https://docs.openwdl.org/
• https://github.com/openwdl/wdl
• https://miniwdl.readthedocs.io/