Linkr

Run a structured quality + security review of recent Linkr changes (security, state-of-the-art code, global coherence, no useless comments, conventions). Logs each review with its date and the commit range covered in REVIEW-LOG.md. Use when the user asks to review the code, audit recent work, or check quality before a release.

Generate a ready-to-import Linkr project ZIP — a project with synthetic dataset(s), a wired dashboard (KPIs + charts + filters), IDE analysis scripts (.py/.r/.sql/.md), and README/tasks, assembled into the importable project ZIP format. Use when the user wants a demo project, sample data, a dashboard, or analysis scripts built from scratch to drag into the app.

Create a new Linkr plugin with R and Python templates. Use this when the user wants to add a new plugin for the Lab (datasets/dashboards) or Warehouse (patient data) scope.

Orchestrates OMOP concept mapping for a Linkr project. Entry point for all mapping work: loads inputs, runs precomputation scripts, and routes to sub-skills (concept-mapping-ai, concept-mapping-drug) based on concept domain. Use when the user wants to map local hospital terminology codes to OMOP standard vocabularies (SNOMED CT, LOINC, UCUM, RxNorm, etc.).

Agentic sub-skill for OMOP concept mapping. Maps source clinical concepts (measurements, conditions, procedures, observations) to OMOP standard concepts using DuckDB search, web search, and clinical reasoning. Called by the concept-mapping orchestrator — do not invoke directly unless you already have a loaded DuckDB session and project context.

Agentic sub-skill for mapping drug concepts to RxNorm Clinical Drug in OMOP. Handles national/local drug terminologies using structured component matching (ingredient + strength + dose form) and ATC/RxNorm relationship traversal. Called by the concept-mapping orchestrator for Drug domain concepts. Do not invoke directly unless you already have a loaded DuckDB session.

Add or update Linkr's bundled "default data" — the seed under apps/web/public/data/seed/ that's pre-loaded into every fresh install (workspaces, projects, datasets, dashboards, databases, ETL pipelines, concept mappings, DQ rule sets, catalogs). Use when the user wants to change the demo/default content, add a seeded project or dataset, bump the bundled MIMIC data, or test the "Default data has been updated" re-seed flow. Covers the unified per-workspace manifest, the id/uid linkage rules, and how to make the user test add/update/remove end to end.

Write and maintain Vitest unit tests for Linkr's pure, critical logic (SQL escaping/validation, OMOP/query builders, fuzzy-search, import/export, format helpers). Defines what to test, what NOT to test, and how to keep tests in sync as the code changes. Use when adding tests, or when you changed tested logic and need to update its test.