Home Mold

implement-galaxy-tool-step

Convert an abstract step into a concrete gxformat2 step using a tool summary.

Mold health

Source layout
Directory Mold with only index.md frontmatter.
Axis fields
target-specific fields are coherent.
Eval plan
Abstract oracle: fixture-independent property checks any run must satisfy.

eval.md declares properties and check type.
eval.md ↗
Scenarios
Concrete cases: fixtures bound to expected values, run against the eval properties.

scenarios.md declares cases bound to fixtures.
scenarios.md ↗
Typed refs
11 typed references; 0 resolver issues.
On-demand triggers
All on-demand references describe triggers.
Evidence checks
Hypothesis references include verification.

axis: target-specific
target: galaxy
name: implement-galaxy-tool-step

contract

Reference Loading

Typed Mold references describe what casting consumes and when the generated skill should load each artifact.

Researchopen-requirements-ledger

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime Upfront Verbatim Hypothesis

Purpose: Carry the open-requirements ledger: read open entries bearing on this step's decisions, mark resolved the ones it closes, and append any new unmet need it surfaces.
Verify: Promote after a worked run shows entries this Mold appends or resolves are consumed downstream without re-derivation.

Schemagalaxy-workflow-draft

Structured contract copied for validation or lookup.

Runtime Upfront Verbatim Cast Validated

Purpose: In/out contract: the draft this Mold reads and mutates in place conforms to [[galaxy-workflow-draft]]. Cast bundles the JSON Schema alongside the [[draft-validate]] CLI checks.

Schemagalaxy-tool-summary

Structured contract copied for validation or lookup.

Runtime Upfront Verbatim Corpus Observed

Purpose: Bind the abstract step against the deterministic tool summary manifest emitted upstream — read `parsed_tool` for ports/datatypes and `input_schemas.workflow_step_linked` for valid step `tool_state` shape.

CLI Commanddraft-validate

Command reference. Usually cast to a sidecar and loaded only when invoked.

Runtime On Demand Sidecar Hypothesis

Purpose: Validate the mutated draft against draft-contract rules; with --concrete, also gate the extracted concrete subset (including the step just implemented) against full gxformat2.
Trigger: After implementing or modifying a concrete tool step in the draft.
Verify: Cast the skill, exercise on an IWC-derived draft with one drafty step, confirm both surfaces validate and diagnostics route back to the implemented step.

Researchgalaxy-workflow-testability-design

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Preserve testable output labels and collection element identifiers while replacing abstract steps with concrete gxformat2 steps.
Trigger: When a concrete step changes output labels, emits collection outputs, creates a diagnostic checkpoint, or makes a final output too weakly assertable.

Researchgalaxy-collection-semantics

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Connect concrete Galaxy tool inputs/outputs while preserving collection mapping and reduction semantics.
Trigger: When implementing a step with data_collection inputs, mapped outputs, reductions, or nested collection wiring.

Researchnextflow-to-galaxy-channel-shape-mapping

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Check whether a concrete tool input/output can preserve the intended source-derived Galaxy collection shape.
Trigger: When implementing concrete steps for source-derived File/list/paired/list:paired/list:list inputs or outputs.

Researchnextflow-operators-to-galaxy-collection-recipes

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Turn operator-derived abstract transforms into concrete Galaxy wiring, collection operations, or review requests.
Trigger: When a concrete step implements behavior traced to map, join, groupTuple, branch, mix, combine, or multiMap.

Researchgalaxy-collection-tools

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Insert built-in Galaxy collection-operation steps when a direct tool connection cannot express the needed shape.
Trigger: When a step needs collection construction, filtering, extraction, zipping, unzipping, flattening, merging, or relabeling.

Researchgalaxy-apply-rules-dsl

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Implement identifier-derived collection reshaping via Apply Rules.
Trigger: When collection element identifiers need regex parsing, nesting-level swaps, regrouping, or paired identifier assignment.

Researchgalaxy-tool-job-failure-reference

Background synthesis loaded by explicit progressive-disclosure metadata.

Runtime On Demand Verbatim Corpus Observed

Purpose: Preserve concrete tool/job failure evidence while implementing step labels, tool ids, output labels, and collection wiring.
Trigger: When a selected wrapper has explicit failure semantics, dynamic outputs, non-default stdio rules, strict-shell behavior, or runtime-only failure risk.

Cast artifacts

Claude skillimplement-galaxy-tool-step— Convert an abstract step into a concrete gxformat2 step using a tool summary.

How to install →

Artifact handoffs

/ pipeline contract

Produces

galaxy-workflow-draft yaml galaxy-workflow-draft.gxwf.yml

gxformat2 skeleton with one more abstract step replaced by a concrete tool step (loop iteration output).
schema

galaxy-workflow-draft

consumed by
- advance-galaxy-draft-step
- repair-galaxy-draft-topology
open-requirements-ledger yaml open-requirements.ledger.yml

Ledger ([[open-requirements-ledger]]) with a blocking entry appended when a step's declared output can't be computed from its wired inputs; drives the fall-through to topology repair.
consumed by

Consumes

galaxy-tool-summary

Galaxy tool summary manifest from [[summarize-galaxy-tool]] conforming to [[galaxy-tool-summary]]; binds the abstract step to a concrete tool's ports via the embedded `parsed_tool` and generated `input_schemas`.
produced by
- summarize-galaxy-tool
galaxy-workflow-draft

gxformat2 skeleton being filled in step by step; the step replaces a placeholder in this draft.
produced by
open-requirements-ledger

Carried obligations ledger [[open-requirements-ledger]]: read entries the design tier recorded against this step before implementing; append a blocking entry if its output can't be computed.
produced by

implement-galaxy-tool-step

Replace one abstract step in the gxformat2 draft with a concrete tool step, using the upstream tool summary. One invocation resolves exactly the chosen step’s TODO_* / _plan_* slots into a concrete tool_id, tool_version, tool_state, and wrapper-determined port names, and returns the mutated draft. This is the “Implement” leaf of the per-step loop owned by advance-galaxy-draft-step.

Single step in scope. This Mold owns the chosen step and the wiring that connects it to ports already in the draft. It does not redesign topology and does not unwind earlier iterations — cross-step rework is the orchestrator’s call.

Sequence

Read the step’s plan. From the galaxy-workflow-draft, take the chosen step’s deferred evidence: _plan_state, _plan_context, _plan_in, _plan_out, and any TODO_* slots the template or data-flow brief left for this phase.
Bind to the tool summary. Read the galaxy-tool-summary manifest: parsed_tool gives concrete input/output port names and datatypes; shape the step’s tool_state against input_schemas.workflow_step_linked. Set tool_version, and set tool_id — confirming or correcting an identity-pinned id rather than re-deriving a good pin from scratch. For a built-in/stock tool the tool_id is the bare id (Filter1, Cut1, collection ops) and tool_version must come from the summary’s cached pin — never invent a stock version; the summary already resolved it against the shed via summarize-galaxy-tool. If input_schemas is null, consult warnings[] for why before binding by hand.
Wire ports. Connect the step’s inputs to their upstream producers and its outputs to downstream consumers per the _plan_in / _plan_out intent, using real wrapper port names. Preserve collection mapping and reduction semantics (galaxy-collection-semantics); for a source-derived shape, check the chosen input/output can actually carry the intended File / list / paired / list:paired shape (nextflow-to-galaxy-channel-shape-mapping).
Close shape gaps. When a direct tool connection cannot express the needed shape, insert a built-in collection-operation step (galaxy-collection-tools); for identifier-derived reshaping — regex parsing, nesting swaps, paired assignment — use Apply Rules (galaxy-apply-rules-dsl); for a transform traced to a Nextflow operator (map, join, groupTuple, branch, mix, combine, multiMap), turn it into concrete wiring or a review request via nextflow-operators-to-galaxy-collection-recipes. A built-in step inserted here is itself a stock tool: resolve its concrete tool_version through the galaxy-tool-cache flow (summarize-galaxy-tool on the bare id) rather than guessing — or leave it draft-tier for the next loop iteration to realize.
Preserve testability. Keep output labels and collection element identifiers stable and addressable (galaxy-workflow-testability-design). Do not rename a labeled output, drop a checkpoint, or make a final output too weakly assertable just to satisfy this step’s wiring.
Validate. Run draft-validate --concrete over the mutated draft: it checks draft-contract rules and gates the extracted concrete subset — including the step just implemented — against full gxformat2. On green, return the draft for the next loop iteration; on red, route the diagnostic back to whichever decision above it implicates.

Failure ownership

When the wrapper can’t cleanly carry what the plan needs — wrong datatype, missing parameter, unsupported collection shape — record where a later failure should be investigated: tool/job failure, data-flow mistake, template wiring mistake, wrapper mismatch, or test/assertion issue. Consult galaxy-tool-job-failure-reference when the selected wrapper has explicit failure semantics (exit-code rules, stdio regex, strict-shell, dynamic outputs); implement the step’s labels and wiring so that evidence survives to runtime rather than being erased by the concretization.

When the step’s declared output cannot be computed from the inputs wired to it — the connection graph says the ports connect, but no wired input carries the evidence the output needs — do not fabricate it. Append a blocking entry to the open-requirements-ledger naming the step, the uncomputable output, and the missing evidence, and fall through to repair-galaxy-draft-topology rather than emitting a step that validates but can’t run. A missing Tool Shed wrapper is not this case; that gap routes through the discover-or-author branch.

Incoming References (65)

advance-galaxy-draft-steprelated mold— Advance the gxformat2 draft by one step: pick the next drafty step, resolve a wrapper, implement the step, and validate.
repair-galaxy-draft-topologyrelated mold— Re-wire a Galaxy draft region when a step's declared output can't be computed from its wired inputs.
Cleanup, sync, and publish non-empty resultsrelated mold— Clean sparse mapped outputs, keep sibling collections aligned, then gate report publishing on non-empty results.
Collection: build list paired with Apply Rulesrelated mold— Use Apply Rules to promote identifier columns into a list:paired collection, with optional cleanup first.
Collection: build named bundlerelated mold— Use BUILD_LIST to assemble named outputs into a collection bundle for publishing or downstream fan-in.
Collection: cleanup after map-over failurerelated mold— Use FILTER_EMPTY or FILTER_FAILED after map-over when bad elements would break downstream collection steps.
Collection: flatten after fan-outrelated mold— Use FLATTEN to collapse nested collection outputs to a flat list once the outer axis no longer matters.
Collection: split identifier via rulesrelated mold— Use Apply Rules regex columns to split one collection identifier into nested list identifiers.
Collection: swap nesting with Apply Rulesrelated mold— Use Apply Rules to regroup a list:list collection by swapping outer and inner identifier columns.
Collection: unbox singletonrelated mold— Use __EXTRACT_DATASET__ with which: first when a one-element collection must become a dataset.
Parameter: compose runtime text parameterrelated mold— Use compose_text_param to build connected text expressions from constants plus runtime scalar values.
Conditional: gate on non-empty resultrelated mold— Derive a boolean from empty or non-empty data, then use when to skip reporting or export steps.
Conditional: route between alternative outputsrelated mold— Use when-gated alternatives plus pick_value to merge binary or one-of-N routes into one downstream value.
Conditional: run optional steprelated mold— Use a workflow boolean connected as inputs.when to skip an optional Galaxy step or branch.
Conditional: transform or pass throughrelated mold— Gate an optional transform, then use pick_value to pass transformed data when present or original data otherwise.
Parameter: derive from filerelated mold— Read a one-value dataset with param_value_from_file, including count recipes that feed typed parameters.
Fan-in bundle, consume, and flattenrelated mold— Bundle parallel outputs into a collection consumer, then flatten nested results for pooled downstream processing.
Galaxy: collection patternsrelated mold— Use this MOC to choose corpus-grounded Galaxy collection transformation patterns.
Galaxy: conditionals patternsrelated mold— Use this MOC to choose corpus-grounded Galaxy when and pick_value conditional patterns.
Galaxy: genomic interval patternsrelated mold— Use this MOC to choose corpus-grounded Galaxy genomic interval operations and recipes on coordinate features.
Galaxy: sequence patternsrelated mold— Use this MOC to choose corpus-grounded Galaxy operations on sequence records (FASTA) — interconvert, reformat, merge, length, extract/mask by region.
Galaxy: tabular patternsrelated mold— Use this MOC to choose corpus-grounded Galaxy tabular transformation patterns.
Collection: harmonize by sortlist from identifiersrelated mold— Use SORTLIST with sort_type:file to reorder one collection by another collection's identifiers.
Interval: consensus regions by multi-intersectrelated mold— Find features reproducible across replicates: multi-intersect per-replicate sets, threshold by replicate count, then intersect back against the merged call.
Interval: compute coveragerelated mold— Two coverage modes: genome-wide depth as a bedgraph (genomecoveragebed) and reads counted in given regions (coveragebed). Same family, different question.
Interval: build a mask by set algebrarelated mold— Compute regions from regions: concatenate candidate intervals, merge into non-overlapping spans, then subtract the set to keep. The gops_* set-algebra recipe.
Interval: merge overlapping featuresrelated mold— Collapse overlapping or book-ended intervals within one set into single spans; bedtools mergebed or the gops_merge Operate-on-Genomic-Intervals tool.
Interval: filter or annotate by overlaprelated mold— Keep, drop, or annotate coordinate features by overlap with a second feature set; bedtools intersect (BED) or vcfvcfintersect (VCF), mapped over a collection.
Interval: window or flank featuresrelated mold— Extend features by a fixed or fractional amount to build neighborhood windows, clamped to chromosome ends; bedtools slopbed with a genome file.
Interval: windowed coverage around featuresrelated mold— Quantify signal in fixed neighborhoods around point features: window the features (slop), collapse overlaps (merge), then count reads in each window (coverage).
Manifest to mapped collection lifecyclerelated mold— Use a manifest or table to build a collection, map a tool per row, then relabel or reshape outputs.
Parameter: map workflow enum to tool parameterrelated mold— Use map_param_value to translate workflow enum values into downstream tool codes, flags, or snippets.
Collection: regex relabel via tabularrelated mold— Derive collection element identifiers in a tabular mapping, then apply them with RELABEL_FROM_FILE.
Sequence: relabel FASTA headers via tabularrelated mold— Edit FASTA headers you cannot easily regex in place: fasta2tab, rewrite column 1 with find/replace, then tab2fasta back. The high-value sequence recipe.
Collection: relabel via rules and find/replacerelated mold— Use Apply Rules, identifier extraction, find/replace, and relabeling for structural fan-out cleanup.
Reshape, relabel, and remap by collection axisrelated mold— Use Apply Rules and deterministic relabeling when domain fan-out creates the wrong map-over axis.
Sequence: compute record lengthsrelated mold— Emit a (id, length) table from a FASTA so downstream tabular steps can filter, sort, or threshold records by length; fasta_compute_length.
Sequence: extract or mask by regionrelated mold— Turn coordinates into sequence: extract FASTA at BED intervals (getfasta), mask regions by BED (maskfasta), or extract transcripts from a GFF (gffread).
Sequence: interconvert FASTA and tabularrelated mold— Move sequence records between FASTA and a (header, sequence) table so tabular tools can edit them; fasta2tab one way, tab2fasta back.
Sequence: merge FASTA and filter uniquerelated mold— Concatenate several FASTA files into one and drop duplicate records by sequence identity in a single step; fasta_merge_files_and_filter_unique_sequences.
Sequence: reformat FASTA line widthrelated mold— Rewrap FASTA records to a fixed sequence-line width so downstream tools and viewers get canonical 60/70/80-column output; cshl_fasta_formatter.
Collection: sync collections by identifierrelated mold— Use collection_element_identifiers with FILTER_FROM_FILE or RELABEL_FROM_FILE to align sibling collections.
Tabular: compute a new columnrelated mold— Use column_maker (Add_a_column1) with strict error_handling to insert/replace a computed column. Per-expression-kind auto_col_types rule.
Tabular: concatenate collection to tablerelated mold— Use collapse_dataset to row-bind a collection of tabulars into one table, with optional element IDs and header dedupe.
Tabular: cut and reorder columnsrelated mold— Use Cut1 with a comma-separated cN list to project — and reorder — columns. Listing out of order is the canonical reorder idiom.
Tabular: filter rows by column valuerelated mold— Use Filter1 with a Python expression over cN columns to drop rows. Highest-frequency tabular row filter in IWC.
Tabular: filter rows by regexrelated mold— Use tp_grep_tool for whole-line regex row filters on tabular input. Grep1 is the legacy alternative.
Tabular: group and aggregaterelated mold— Use datamash_ops for grouped tabular aggregation: multi-column grouping, collapse, countunique, min/max, and reductions.
Tabular: join on keyrelated mold— Use tp_easyjoin_tool for two-tabular key joins; use tp_multijoin_tool for many files and query_tabular for SQL joins.
Tabular: pivot collection to widerelated mold— Use collection_column_join to outer-join a collection of 2-column id/value tables into one wide table.
Tabular: prepend headerrelated mold— Use tp_awk_tool to prepend a constant header line, optionally skipping or reformatting an existing first row.
Tabular: relabel by row counterrelated mold— Use tp_awk_tool to replace each row or label with deterministic sample_N values from awk NR.
Tabular: split taxonomy stringrelated mold— Use tp_awk_tool to split semicolon-delimited taxonomy strings into explicit rank columns with missing-rank handling.
Tabular: SQL queryrelated mold— Use query_tabular when SQL semantics justify it: windows, joins, anti-joins, or fused project+compute over tabulars.
Tabular: synthesize BED from 3-column inputrelated mold— Use tp_awk_tool to convert chrom/start/end rows into 6-column BED, subtracting 1 from start and setting constants.
Tabular: to collection by rowrelated mold— Use split_file_to_collection split_by:col to fan a tabular into collection elements by row/key.
Galaxy tool and job failure referencerelated mold— Reference for Galaxy tool stdio rules, job failure detection, job states, and job API failure surfaces.
Galaxy workflow draft formatrelated mold— gxformat2 draft superset: wrapper-tier TODOs (tool_id, tool_state, port names) plus _plan_state / _plan_context / _plan_in / _plan_out per tool step.
Nextflow operators to Galaxy collection recipesrelated mold— Classifies common Nextflow operators as Galaxy wiring, collection semantics, explicit steps, or review triggers.
Nextflow-to-Galaxy channel shape mappingrelated mold— Maps common Nextflow channel, tuple, and path shapes to Galaxy dataset and collection shapes.
Open-requirements ledgerrelated mold— Carried unresolved-requirements artifact the source→Galaxy pipeline discharges or explicitly surrenders, autonomously.
Galaxy tool discovery recommendationrelated mold— JSON Schema for Tool Shed discovery hit, weak, and miss recommendations.
Galaxy tool summary manifestrelated mold— JSON Schema for the deterministic per-tool manifest emitted by `galaxy-tool-cache summarize`.
Galaxy workflow draft (gxformat2 superset)related mold— JSON Schema for `class: GalaxyWorkflowDraft` — gxformat2 with `TODO_*` sentinels and `_plan_*` planning fields per draft step.
Galaxy ParsedToolrelated mold— JSON Schema for the upstream Galaxy `ParsedTool` model, vendored from `@galaxy-tool-util/schema`.