Foundry
Home pipeline · nextflow → cwl

NEXTFLOW → CWL

Direct path from a Nextflow pipeline to a CWL Workflow + CommandLineTool set.

draft source/nextflowtarget/cwl

Phases

/ subway map 12 phases
  1. summarize-nextflow
  2. nextflow-summary-to-cwl-interface
  3. nextflow-summary-to-cwl-data-flow
  4. summary-to-cwl-template
  5. summarize-cwl-tool loop
  6. implement-cwl-tool-step loop
  7. validate-cwl loop
  8. nextflow-test-to-cwl-test-plan
  9. implement-cwl-workflow-test
  10. validate-cwl
  11. run-workflow-test
  12. debug-cwl-workflow-output

Run this pipeline

/ stop-gap harness

Run this pipeline today via its assembled harness skill pipeline-nextflow-to-cwl — install once (run guide), then select it in your runtime: 

Claude:      /foundry-skills:pipeline-nextflow-to-cwl
Codex:       $pipeline-nextflow-to-cwl
Antigravity: /pipeline-nextflow-to-cwl

Each run gets its own ./<run-slug>/ working directory so artifacts stay namespaced and a second run won't clobber the first — the harness sets this up first (phase detail).

phases
12
auto casts
12
manual
0
loops
3

NEXTFLOW → CWL

Start with a Nextflow source tree pinned to a tag or commit and its available nf-test cases, fixtures, snapshots, and expected outputs. Review the selected test case in summary-nextflow.json. If no whole-pipeline case is available, record the narrower scope and deferred cases before planning the CWL test.

Review the CWL interface and data-flow briefs against the source. Confirm primary inputs and outputs, scatter and when choices, and any stated scope reductions. For each placeholder step, check the command, software provenance, and input and output bindings before implementing and validating it. The CWL loop has no shared completion check, so inspect the draft for remaining placeholders and unresolved run: targets.

Check that cwl-test-plan.md and cwl-job.yml use the selected source test evidence and match the implemented CWL ports. Report unavailable fixtures or narrow coverage instead of treating a partial test as full coverage.

Inspect the final validation diagnostics and any workflow-test-result.json or CWL debug report. Structural validation does not establish scientific equivalence. The CWL run path remains unproven end to end, so report whether a test ran, what it exercised, and any failures with the workflow.