Galaxy Workflow Foundry
Working Claim
Runtime retrieval is not enough for schema-bound workflow construction. Foundry treats skills as compiled artifacts from a curated knowledge base, with schemas and provenance making the generated instructions inspectable and auditable.
Current Emphasis
This is the most speculative paper. It needs a concrete case study before it becomes a conventional software manuscript. The demonstrated path is construction from intent (interview-to-galaxy); cross-source conversion is architecturally supported but not yet exercised end-to-end. Because the construction path supplies no demonstrated biological application, the working venue has been pulled back from Genome Biology (see Target Ladder).
Target Ladder
Aim high but maintain a credible retreat path. Draft once at the highest realistic target’s length; fallbacks are trim operations on that draft, not rewrites. Pulled back one rung (2026-06): the demonstrated path is construction-from-intent with no biological-application story, so the working primary is Genome Research, not Genome Biology.
Working primary — Genome Research, Methods/Resource, ~6000–8000 words. Infrastructure-with-applied-validation framing, with Planemo precedent at the venue. Carries the construction case study (gxwf- and planemo-validated, provenance-traced) plus the failure-comparison vignette against an unguided-agent baseline. Does not require demonstrated biological discovery — the contribution is the compiler/typed-draft/provenance machinery, validated on a real construction task.
Aspirational — Genome Biology, Method/Software article, ~6000–8000 words. Gated on capability the project does not have today. Genome Biology’s Software section expects demonstrated biological application; that needs at least 2 worked case studies on real published pipelines (recovered signal, parameter-drift caught, reproducibility delta) — which means the conversion paths working end-to-end, not just construction-from-intent. Keep the door open only if a PhD-contributor lands a real conversion + biology result; do not plan the draft around it.
Fallback 2 — Bioinformatics Original Paper, ~5000 words. Trim from the higher draft: one case study instead of N, schema-as-contract pattern compressed to one section, drop comparison-to-runtime-retrieval framing, treat Molds as a single concept rather than a typology. Loses the “compiler” framing’s depth; reduces to “agent-authoring pipeline for Galaxy workflows backed by gxwf validation.”
Fallback 3 — Bioinformatics Application Note, ~2000 words. Single case study, single Mold pipeline, scaffolding-level description only. Submit only as a citation-availability artifact while a longer version is in revision elsewhere; not a credible standalone home for this contribution.
Off-ladder alternatives. Nature Methods is possible if a case study uncovers a publishable methodological error or recovers a non-trivial biological result; do not plan for this, but leave the door open if a PhD contribution lands one. PLOS Computational Biology if the agent-authoring framing becomes the dominant story.