Engine skill — Nextflow

[oesteban]

Objective

Build a grounded skill for Nextflow so the agent emits/compiles a reconstruction targeting it from grounded knowledge (the KB + this skill) rather than parametric recall of Nextflow syntax. The skill grounds Nextflow's DSL2 model — processes, typed channels and dataflow composition, modules, executors (local / SLURM / Kubernetes), containerization, and resumability. The skill is validated by reconstructing the simple anatomical workflow established in #14 (brain extraction → INU correction → tissue segmentation → spatial normalization), executed on the anatomical (T1w) images of a subset of the OpenNeuro testbench (nipreps/fmriprep-equivalence#3).

Rationale

Following the logic of the earlier workflow skills, Nextflow is the evident next skillable target: it can be verified against the CWL output of #16 (Nextflow executes CWL), giving the reconstruction an independent cross-check. Nextflow is the project's primary static-engine target and underpins the full Nextflow reconstruction (#7).

The transplant thesis holds only if a reconstruction is grounded: emitting Nextflow from parametric memory would be unverifiable, reproducing the "did the model just memorize it" failure mode the project rejects. A grounded skill makes every emitted construct traceable to an authoritative description of the engine.

Methodological approach

Important

This is a separate skill family from niskiller (nipreps/skills-comm#6): Nextflow is not a neuroimaging tool, and the boundary is structural — engine skills live in agentic-files, neuroimaging-tool skills in skills-comm.

1. Author the skill. A lean decision surface plus a grounded reference of Nextflow's execution model, each construct tied to an authoritative source (official docs / reference) with provenance (version, retrieval date).
2. Ground the lowering. How a KB operation lowers to Nextflow constructs, so emission is a grounded transformation.
3. Validate. Reconstruct the Engine skill — Claude Code /workflow #14 simple anatomical workflow on Nextflow, execute it on the testbench T1w subset (Assemble the fMRIPrep testbench dataset drawing from OpenNeuro fmriprep-equivalence#3), and assess reliability with Bland–Altman analyses against the Nipype reference of the same workflow (Nipype baseline workflow niworkflows#1053), and cross-check against the CWL specification from Engine skill — CWL #16.

Outcomes/impact

1. A grounded Nextflow skill.
2. The Engine skill — Claude Code /workflow #14 anatomical workflow reconstructed on Nextflow, validated (Bland–Altman vs the Nipype reference) on the testbench T1w subset.

Impact. The primary static engine, cross-validated against CWL — underpinning the full Nextflow reconstruction (#7).

Direct upstream issues

(issues this one depends on / builds upon)

• Engine skill — CWL #16
• Nipype baseline workflow niworkflows#1053

Direct downstream issues

(issues that depend on this one)

• Workflow Architect Agent 1.0 #1