irw

Use this skill when asked to regenerate or refresh the IRW dictionary, metadata, tags, or biblio CSVs that feed the Redivis "irw_meta:bdxt" dataset (metadata.csv, biblio.csv, tags.csv, comps_metadata.csv, nominal_metadata.csv, simsyn_metadata.csv, comps/nominal/simsyn biblio.csv, itemtext_metadata.csv, hero_stats.json), to audit/reconcile table names across the metadata/tags/biblio tables and the live Redivis IRW datasets, or to actually upload those regenerated CSVs into the Redivis irw_meta tables. Also applies to phrases like "run the metadata pipeline", "update Redivis metadata", "check for table name mismatches", "add item_response_warehouse_3 to metadata", "which tables are missing from the dictionary/tags/biblio", or "upload biblio/tags/metadata to Redivis".

This skill should be used when the user asks to "find new datasets for IRW", "run discovery", "search for item response datasets", "triage candidates", "retriage human_assistance rows", "process the queue", or otherwise references the automated_finding pipeline (irw_discover_updated.py, irw_batch_updated.py, irw_retriage_ha.py, irw_discover_plos.py, irw_process_queue.py) or its TODO.md/BATCH_LOG.md. Also applies to searching individual open-access journals (e.g. PLOS ONE) directly, as opposed to data repositories.

Use this skill when asked to create a new IRW vignette — a Quarto analysis page for itemresponsewarehouse.org — scaffolding the compute script, running a pilot, and writing the .qmd page following the site's established compute/render split and canonical scaffold. Applies to phrases like "start a new vignette", "scaffold a vignette for X", "write a vignette replicating <paper>", or "assumption-audit vignette for Y". Does NOT cover editing vignettes/index.qmd, navbar/_quarto.yml changes, or merging to main — those are explicitly out of scope here.

This skill should be used when the user asks to "tag this table", "process the tagging queue", "add IRW tags for X", or otherwise references filling in the "IRW Tags" Google Sheet (columns table/Rater/Construct Name/Context Text/Item text available?/Age Range/Child Age/Sample/Construct type/Measurement tool/Item format/Primary Language(s)/Notes) for newly added IRW tables. Also applies when asked to re-derive or check the sheet's controlled vocabularies.

This skill should be used when the user asks to "generate item text for X", "process the itemtext queue", "extract items for this table", or otherwise references transcribing/extracting instrument, section, item, or response-option text for an IRW table from its source paper. Also applies when the user references itemtext/join.R, itemtext/upload.py, the itemtext index workbook, or the itemtext.html public schema.