Objective
Create a NiQuery Agent Skill, built with Claude's /skill-creator plugin, that makes NiQuery — NiPreps' principled tool for querying open neuroimaging datasets — operable by an agent. The skill wraps NiQuery's query surface so an agent can programmatically search OpenNeuro (Markiewicz et al., 2021) for datasets matching precise acquisition and metadata criteria, extract the relevant information, and assemble curated subsets.
Rationale
Assembling a benchmark that exercises every fMRIPrep compute-graph branch (nipreps/fmriprep-equivalence#3) means finding OpenNeuro datasets with specific properties — fieldmaps present/absent and by estimation type, single- vs multi-echo, single- vs multi-run/-session, slice-timing metadata present/absent, surface- vs volume-eligible — across thousands of candidate datasets. Doing that by hand does not scale and is not reproducible. A NiQuery skill turns that search into a grounded, repeatable agent capability, and is the querying instrument the testbench-assembly issue depends on.
Methodological approach
- Author with
/skill-creator. Use the /skill-creator plugin to generate the skill from NiQuery's CLI/API surface, capturing purpose, inputs/outputs, query idioms, and pitfalls.
- Target OpenNeuro querying. Cover the queries the benchmark needs — filtering by BIDS entities and metadata (echoes, fieldmaps, sessions/runs, slice-timing, task/rest) — and resolve matches to version-pinned snapshots.
- Validate. Exercise the skill against known OpenNeuro datasets, checking that queries return the expected matches and that results are reproducible.
Outcomes/impact
- A NiQuery skill that lets an agent query OpenNeuro for datasets by acquisition/metadata criteria and assemble curated subsets.
Impact. The skill is the querying instrument that makes the OpenNeuro testbench (nipreps/fmriprep-equivalence#3) assemblable programmatically and reproducibly, rather than by manual dataset triage.
Direct upstream issues
(issues this one depends on / builds upon)
None — wraps the existing NiQuery software (nipreps/niquery).
Direct downstream issues
(issues that depend on this one)
Objective
Create a NiQuery Agent Skill, built with Claude's
/skill-creatorplugin, that makes NiQuery — NiPreps' principled tool for querying open neuroimaging datasets — operable by an agent. The skill wraps NiQuery's query surface so an agent can programmatically search OpenNeuro (Markiewicz et al., 2021) for datasets matching precise acquisition and metadata criteria, extract the relevant information, and assemble curated subsets.Rationale
Assembling a benchmark that exercises every fMRIPrep compute-graph branch (nipreps/fmriprep-equivalence#3) means finding OpenNeuro datasets with specific properties — fieldmaps present/absent and by estimation type, single- vs multi-echo, single- vs multi-run/-session, slice-timing metadata present/absent, surface- vs volume-eligible — across thousands of candidate datasets. Doing that by hand does not scale and is not reproducible. A NiQuery skill turns that search into a grounded, repeatable agent capability, and is the querying instrument the testbench-assembly issue depends on.
Methodological approach
/skill-creator. Use the/skill-creatorplugin to generate the skill from NiQuery's CLI/API surface, capturing purpose, inputs/outputs, query idioms, and pitfalls.Outcomes/impact
Impact. The skill is the querying instrument that makes the OpenNeuro testbench (nipreps/fmriprep-equivalence#3) assemblable programmatically and reproducibly, rather than by manual dataset triage.
Direct upstream issues
(issues this one depends on / builds upon)
None — wraps the existing NiQuery software (
nipreps/niquery).Direct downstream issues
(issues that depend on this one)