Harness Pipelines

Ordered Mold sequences.

Pipelines compose Molds into harness journeys and make the Foundry navigable from source to target.

Run a pipeline → — how to run one end-to-end today.

Total
7
Phases
81
PipelineSummaryPhasesStatus
CWL → GALAXY Path from a CWL Workflow to a Galaxy gxformat2 workflow. Lighter upstream extraction. 12 draft
INTERVIEW → GALAXY Interview-driven path to a Galaxy gxformat2 workflow through the shared freeform-summary handoff. 12 draft
NEXTFLOW → CWL Direct path from a Nextflow pipeline to a CWL Workflow + CommandLineTool set. 12 draft
NEXTFLOW → GALAXY Direct path from a Nextflow pipeline to a Galaxy gxformat2 workflow. 13 draft
PAPER → CWL Direct path from a paper to a CWL Workflow + CommandLineTool set. 11 draft
PAPER → GALAXY Direct path from a paper to a Galaxy gxformat2 workflow. No CWL intermediate. 12 draft
UPDATE-INTERVIEW → GALAXY Interview-driven, edit-in-place modification of an existing Galaxy gxformat2 workflow via a reviewable change-set, reusing the per-step draft loop. 9 draft

Run a pipeline

/ how to run one end-to-end today

A Foundry pipeline is an ordered sequence of Molds that walks a source workflow to a target workflow. You run one today via its assembled pipeline-* harness skill — a stop-gap that invokes each phase's cast in order.

  1. Install the harness skills. They install like any Foundry cast skill through the Claude Code or Codex plugin metadata around the same shared bundle. See Install.
  2. Make the runtime available. Pipelines call out to gxwf (schema/validation), Planemo (test execution), and the Foundry validators. Install these in the environment before a run that reaches the validate/test phases.
  3. Invoke the harness for the pipeline you want. Each run gets its own ./<run-slug>/ working directory, so artifacts stay namespaced and a second run won't clobber the first — the harness creates it before phase 1.
  4. Optional run flags. Pass --use-subagents to run each cast phase in its own subagent (one per loop iteration) so the orchestrator's context stays small, and --checkpoint to git init the run directory and commit after every phase — a per-step history for workflow-implementation visualizations. Both default off and compose.

See the 7 runnable harnesses on Usage → — invoke command, phase counts, and install steps.