← all publishers

nealcaren

@nealcaren source repo

19 published skills

  1. Sample User · nealcaren
    Run a two-agent "sample user" usability audit of a topica model's end-to-end analysis workflow, from the seat of a first-time domain-expert researcher. Use when a user wants to evaluate how usable a model is in practice ("do a sample-user run on LDA", "sample-user the new keyATM covariate path", "have sample users try STM"), or as the standard post-merge acceptance gate for any newly added model. Dispatches one Claude and one Gemini agent, each role-playing a first-time computational social scientist who runs the whole workflow (vocabulary -> choosing K -> fit -> robustness -> validation -> covariate effects -> reporting) on a bundled sample dataset, then synthesizes their friction into a ranked, verified GitHub issue. Complements add-topic-model (which ships a model); this audits whether a real user can actually use it.
    0
    installs
  2. Add Topic Model · nealcaren bundle
    End-to-end workflow for adding a new topic-modeling algorithm to the topica library (a Rust + PyO3 + numpy package). Use when a user asks to add, port, or implement a topic model (e.g. "add CTM-2", "port BTM", "implement an anchored model") into topica. Covers grounding in the literature, finding and running the reference implementation to build a gold-standard result set, implementing a Rust core with Python bindings under topica's conventions, a two-reviewer dual-review gate (Codex + Gemini, one faithful-parity and one adversarial) run both after the plan is approved and again before the PR, benchmarking against the reference, and shipping via a GitHub issue + PR that also updates the README and docs.
    0
    installs
  3. Topica Analysis · nealcaren
    Working guide for an LLM agent helping a social scientist run a topic-modeling analysis with the topica Python library — building a defensible corpus, choosing and justifying K, fitting a model, validating it, estimating covariate effects with honest uncertainty, and reporting results. Use when the user wants to USE topica to analyze a text corpus (not extend the library). The companion developer skill for adding a model is add-topic-model.
    0
    installs
  4. Bibliography Builder · nealcaren bundle
    Bibliography Builder
    0
    installs
  5. Tidy R · nealcaren bundle
    Modern tidyverse patterns, style guide, and migration guidance for R development. Use this skill when writing R code, reviewing tidyverse code, updating legacy R code, or enforcing consistent style. Covers native pipe usage, join_by() syntax, .by grouping, pick/across/reframe, filter_out/when_any/when_all, recode_values/replace_values/replace_when, tidyselect helpers, .data/.env pronouns, stringr, naming conventions, and readr.
    0
    installs
  6. Review R · nealcaren
    Read-only code review for R analysis scripts. Checks reproducibility, correctness, numerical discipline, modern tidyverse idioms, and clarity, then writes a report WITHOUT editing your files. Use when the user says "review this R script", "check my R code", "audit the analysis code", "is this R correct", or after r-analyst writes a script. Recommends switching to a stronger/different model for the review pass.
    0
    installs
  7. Review Py · nealcaren
    Read-only code review for Python analysis scripts and notebooks (pandas, statsmodels, scikit-learn, and text-analysis pipelines). Checks reproducibility, correctness, pandas/NumPy numerical traps, data leakage, and whether reported numbers match what the code produces — then writes a report WITHOUT editing your files. Use when the user says "review this Python script", "check my notebook", "audit the analysis", or after text-analyst writes code.
    0
    installs
  8. Stat Check · nealcaren bundle
    Guard against hallucinated statistics in write-ups. Defines the results-ledger convention (analysis scripts emit every reportable number to output/results.json at full precision) and runs a reconciliation pass that flags any number in a report's prose that doesn't trace back to a computed value ("orphans"). Use after drafting a Results/findings memo or report, before finalizing, or whenever the user asks to check that reported numbers match the analysis. Invoked by r-analyst/stata-analyst Phase 5, the review skills, and repro-package.
    0
    installs
  9. Method Scout · nealcaren bundle
    Protocol for adopting a statistical method that isn't in the analysis skills' built-in technique menu (e.g. latent class analysis, mixture models, SEM, survival, sequence analysis, GEE). Instead of guessing an API from memory, it finds the maintained, canonical package via the ecosystem's own index, reads and follows the package's vignette, applies the same identification/validation/reproducibility discipline, and benchmarks against the vignette before touching real data. Use when a user asks for a method r-analyst/stata-analyst/text-analyst don't already have a technique guide for, or says "how do I do <method> in R/Stata/Python".
    0
    installs
  10. Review Stata · nealcaren
    Read-only code review for Stata .do files. Checks reproducibility, correctness, Stata's notorious numerical/missing-value traps, idioms, and whether reported numbers match what the code produces — then writes a report WITHOUT editing your files. Use when the user says "review this do-file", "check my Stata code", "audit the analysis", or after stata-analyst writes a script.
    0
    installs
  11. Repro Package · nealcaren bundle
    Turn a finished analysis into a submission-ready replication archive — prove it rebuilds from raw in a clean session, then assemble the master runner, README, codebook, environment snapshot, data-access statement, and an output→exhibit map that journals now require. Works for R, Stata, or Python projects. Use when the analysis is done and the user is preparing to submit, deposit replication files, or says "make a replication package", "build the reproducibility archive", "get this ready for submission".
    0
    installs
  12. Data Acquisition · nealcaren bundle
    Acquire the data a quantitative sociology analysis starts from — Dataverse replication packages, IPUMS/Census (ACS), GSS, ANES, ICPSR, OSF, and generic archives — and land it as immutable, provenanced raw data with a codebook and a re-runnable fetch script. Use at the very start of a quantitative project, before r-analyst / stata-analyst. Triggers on "get the replication data for this paper", "download this Dataverse dataset", "pull ACS / GSS / IPUMS", "grab the data from this DOI/OSF link".
    0
    installs
  13. Project Scaffold · nealcaren bundle
    Initialize research project structure with standardized directories and metadata files. Creates project.yaml (configuration) and progress.yaml (state tracking) that other skills use for canonical paths and workflow coordination. Supports qualitative, quantitative, and mixed methods projects. Use when starting a new project or adopting an existing one.
    0
    installs
  14. Unc Hpc · nealcaren bundle
    Run computational jobs on UNC Chapel Hill's HPC clusters (Longleaf and Sycamore). Use when a user wants to: submit SLURM batch jobs, set up Python/conda environments for ML/deep learning, run PyTorch or TensorFlow training on GPUs, run Hugging Face model inference or fine-tuning (including Unsloth for fast QLoRA), use Jupyter notebooks on the cluster, transfer data, manage storage, monitor GPU utilization, write SLURM scripts, do hyperparameter sweeps, or anything involving UNC Research Computing infrastructure. Covers both Longleaf (high-throughput, broad GPU selection) and Sycamore (HPC, H100 GPUs).
    0
    installs
  15. Lit Search · nealcaren bundle
    Build systematic literature databases for sociology research using OpenAlex API. Guides you through search, screening, snowballing, annotation, and synthesis with structured user interaction at each stage.
    0
    installs
  16. Text Analyst · nealcaren bundle
    Computational text analysis for sociology research using R or Python. Guides you through topic models, sentiment analysis, classification, and embeddings with systematic validation. Supports both traditional (LDA, STM) and neural (BERT, BERTopic) methods.
    0
    installs
  17. Stata Analyst · nealcaren bundle
    Stata statistical analysis for publication-ready sociology research. Guides you through phased workflows for DiD, IV, matching, panel methods, and more. Use when doing quantitative analysis in Stata for academic papers.
    0
    installs
  18. Interview Analyst · nealcaren bundle
    Pragmatic qualitative analysis for interview data in sociology research. Guides you through systematic coding, interpretation, and synthesis with quality checkpoints. Supports theory-informed (Track A) or data-first (Track B) approaches.
    0
    installs
  19. R Analyst · nealcaren bundle
    R statistical analysis for publication-ready sociology research. Guides you through phased workflows for DiD, IV, matching, panel methods, and more. Use when doing quantitative analysis in R for academic papers.
    0
    installs