kenlkehl

kenlkehl/onc-data-wrangler-plugin

Plugin Claude Code

Plugin marketplace listing 1 plugin: onc-data-wrangler.

kenlkehl/onc-data-wrangler-plugin

Plugin Claude Code

Oncology data wrangling: extract structured data from clinical notes, build privacy-preserving DuckDB databases, query cohorts, reproduce published paper results, and perform individual-level analysis.

kenlkehl/onc-data-wrangler-plugin

Agent

Per-question analysis worker. Forwards a single research question to the analyze-data skill in answerone mode and exits. Exists as a separate agent solely to enforce Phase-2 blinding in reproduce-paper: the wrapper's context is scoped to the question + data paths and cannot see paper PDFs or ground-truth answers.…

kenlkehl/onc-data-wrangler-plugin

Agent

Per-document clinical-note compression worker. Summarizes one clinical document into a concise oncology-focused summary and writes JSON output. Spawned by the compress-notes skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Data file validation worker. Inspects downloaded data files for a single paper, determines if they contain analyzable tabular data. Writes structured JSON result to a specified output path. Spawned by the pull-papers skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Per-row discrepancy worker for reproduce-paper Phase 3. Forwards a (reportedresult, modelresult, question, paperpdf) payload to the analyze-data skill in compare mode and exits. Phase 3 is allowed to see the paper PDF -- the blinding boundary only applies to Phase 2. Spawned by the reproduce-paper skill orchestrator…

kenlkehl/onc-data-wrangler-plugin

Agent

Generates a single synthetic clinical document (clinical note, imaging report, pathology report, or NGS report) for one event in a patient's timeline. Uses the masked-text approach with full patient event context. Spawned by the generate-synthetic-data skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Generates a synthetic patient event list for one patient matching a clinical scenario. Writes structured JSON output (patientid, events array) to a specified path. Spawned by the generate-synthetic-data skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Per-patient clinical note extraction worker. Reads ontology definitions and extracts structured data from clinical text using domain-group-based extraction. Writes structured JSON result to a specified output path. Spawned by the extract-notes skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Paper question extraction worker. Reads a published biomedical paper PDF, cross-references with data dictionaries, and extracts every quantitative result as an independently answerable analysis question. Writes structured output files (questionswithanswers.xlsx, questionsonly.xlsx, papercontext.txt). Spawned by the…

kenlkehl/onc-data-wrangler-plugin

Agent

Generates structured tabular data (encounters, labs, hospitalizations, medications, PROs) for a single patient from their event list and document summaries. Reads table schemas from YAML files. Writes JSON output to a specified path. Spawned by the generate-synthetic-data skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Post-extraction cross-field validation worker. Loads extraction results for a batch of patients, runs NAACCR cross-field edit checks and confidence scoring, and produces a review queue. Spawned by the extract-notes skill -- do not invoke directly.

kenlkehl/onc-data-wrangler-plugin

Agent

Per-patient verification worker for the verify-answers skill. Given one patient's QA-extraction answer dict, performs a three-tier check (evidence-consistency, cross-answer consistency, targeted retrieval for flagged cells) and writes a verified answer dict plus a per-cell audit trail to a specified output path.…

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Interactive querying of the project's DuckDB database. Supports aggregate queries (with privacy enforcement) and individual-level queries (when privacy mode allows). Use when the user wants to explore or analyze data in the built database.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Interactive Python-based data analysis for oncology datasets. Supports ad-hoc analysis with pandas, survival analysis, statistical modeling, and deep oncology/biostatistics domain knowledge. Works with DuckDB databases or raw data files. Use when the user wants to explore or analyze data using Python rather than SQL.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Answer clinical questions about patients from their notes using LLM-based extraction. Provide a questions file (one question per line) and a notes file to get per-patient answers with confidence scores and evidence. Use when the user wants to answer specific clinical questions across a patient cohort.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Create a custom ontology definition from a data dictionary or codebook. Generates YAML ontology files that can be used for extraction. Use when the user has a data dictionary and wants to define extraction fields.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Summarize individual oncology clinical documents into concise document-level summaries using local OpenAI-compatible/vLLM, Azure OpenAI, Anthropic, Vertex Claude, Gemini, or Claude Code native mode. Use when the user wants compressed clinical notes or short per-document clinical summaries before extraction, review, or…

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

De-identify one structured tabular data file (CSV, TSV, or parquet). Detects likely PHI columns, replaces patient IDs/MRNs/names with stable pseudonyms and realistic fake names, shifts dates per patient, and optionally uses an explicitly approved LLM to rewrite short free-text clinical evidence columns.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Create a one-row-per-patient final analysis dataset from a DuckDB database or raw tabular files. Interactive column definition with oncology/biostatistics guidance, progressive previews, and reproducible script generation. Use when the user wants to build an analytic dataset for statistical modeling or export.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Extract structured data from unstructured clinical notes using dictionary-driven LLM extraction. Supports local models, Azure, Claude API, Google Gemini, or Claude Code native extraction. Use when the user wants to extract data from clinical text.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Generate synthetic clinical data (patient events, clinical documents, and structured tables) from one or more clinical scenario descriptions. Supports external LLM backends with parallel Python workers, or Claude Code native generation with parallel subagents.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Interactively build a DuckDB database from raw tabular data files. Discovers files, configures the project, builds cohort, loads structured data, and creates a queryable privacy-preserving database. Use when the user wants to go from raw data files to a DuckDB database.

kenlkehl/onc-data-wrangler-plugin

Skill Claude CodeCodex

Search PubMed Central for oncology papers with analyzable data, classify by research category (basic science, computational biology, translational, clinical), validate data quality by inspecting files, and organize into category folders. Use when the user wants to find and download oncology research papers with data…