joshzyj

joshzyj/open-scholar-skill

Plugin Claude Code

Open Scholar — AI-assisted research skills for the social sciences, from research question to submission-ready manuscript and replication package.

joshzyj/open-scholar-skill

Plugin Claude Code

AI-assisted academic paper writing for social sciences. 36 skills + 20 specialized agents (9 peer-reviewer + 6 code-review + 5 verification) covering the research pipeline: idea formalization (5-agent evaluation panel), systematic literature review (PRISMA 2020), theory and hypothesis development (25+ frameworks)…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in computational social science methods — NLP/text-as-data, machine learning, network analysis, LLM annotation, computer vision, agent-based modeling, and computational reproducibility. Invoked by scholar-respond to generate a methods-focused review of computational papers.…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in population representativeness, demographic measurement, and population-level inference. Invoked by scholar-respond to generate a demographics-focused review of a social science manuscript. Evaluates sampling strategy, demographic composition, lifecycle analysis, decomposition…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in research ethics, IRB compliance, informed consent, vulnerable populations, data privacy, and responsible conduct of research. Invoked by scholar-respond to generate an ethics-focused review of a social science manuscript. Evaluates IRB approval, consent procedures…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in sociolinguistics, discourse analysis, and language variation — variationist methods, conversation analysis, interactional sociolinguistics, language contact, and computational linguistics. Invoked by scholar-respond to generate a methods-focused review of linguistics and…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in mixed-methods research design, qualitative-quantitative integration, and methodological coherence across study components. Invoked by scholar-respond to generate a mixed-methods-focused review of a social science manuscript. Evaluates integration strategy, alignment of…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in qualitative and ethnographic methods — in-depth interviews, ethnography, discourse analysis, grounded theory, and mixed methods. Invoked by scholar-respond to generate a methods-focused review of qualitative social science manuscripts. Evaluates methodological transparency…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in quantitative methods, causal inference, and empirical rigor. Invoked by scholar-respond to generate a methods-focused review of a social science manuscript. Evaluates research design, identification strategy, model specification, robustness, and data quality.

joshzyj/open-scholar-skill

Agent Claude Code

A simulated senior peer reviewer who provides a holistic, editorial-level evaluation of a social science manuscript. Invoked by scholar-respond to assess overall significance, suitability for the target journal, writing quality, framing, and to issue an editorial recommendation. Acts as the "third reviewer" and…

joshzyj/open-scholar-skill

Agent Claude Code

A simulated peer reviewer specializing in sociological theory, conceptual frameworks, and the theoretical contribution of empirical papers. Invoked by scholar-respond to generate a theory-focused review. Evaluates theoretical coherence, hypothesis derivation, literature engagement, and the paper's contribution to…

joshzyj/open-scholar-skill

Agent Claude Code

A code review agent that checks analysis scripts for logical errors, incorrect function usage, wrong variable references, off-by-one errors, silent coercion bugs, and data manipulation mistakes that could produce wrong results. Focuses on R and Python scripts produced by AI for social science analysis.

joshzyj/open-scholar-skill

Agent Claude Code

A code review agent that verifies variable construction, recoding, categorization, sample restrictions, and data transformations against codebooks, data dictionaries, and design documents. Catches miscoded categories, wrong value labels, reversed scales, incorrect aggregation, and mishandled missing value codes in…

joshzyj/open-scholar-skill

Agent Claude Code

A code review agent that evaluates whether analysis scripts form a complete, self-contained, and reproducible pipeline — checking dependency management, file path portability, execution order, environment specification, and documentation sufficient for independent replication.

joshzyj/open-scholar-skill

Agent Claude Code

A code review agent that checks analysis scripts for fragile patterns, missing error handling at data boundaries, hardcoded assumptions, edge cases, and silent failures that could break under different data conditions. Focuses on defensive coding for social science research pipelines.

joshzyj/open-scholar-skill

Agent Claude Code

A code review agent that verifies the statistical methodology implemented in scripts matches the research design specification — correct estimator, identification strategy, standard errors, hypothesis tests, effect size computation, and reporting standards for social science journals.

joshzyj/open-scholar-skill

Agent Claude Code

A code review agent that evaluates code quality, readability, naming conventions, DRY violations, dead code, and maintainability of AI-generated analysis scripts. Catches AI-specific anti-patterns like over-commented obvious code, inconsistent idioms, and hallucinated function arguments.

joshzyj/open-scholar-skill

Agent Claude Code

A verification agent that independently checks, at the sentence level, whether each cited source actually SUPPORTS the prose claim attributed to it — distinct from whether the reference exists. For every (claim sentence, citation) pair it decomposes the sentence into atomic sub-claims (direction, magnitude…

joshzyj/open-scholar-skill

Agent Claude Code

A verification agent that ensures all artifacts (raw outputs, manuscript tables, manuscript figures) are properly cross-referenced, sequentially numbered, and complete. Checks that every raw output has a manuscript counterpart, every manuscript table/figure is referenced in text, and no orphaned or missing artifacts…

joshzyj/open-scholar-skill

Agent Claude Code

A verification agent that performs Stage 1 figure verification — comparing raw figure outputs (PDFs, PNGs from analysis scripts) against figure descriptions and captions in the manuscript, and checking that the data underlying each figure is consistent with the raw analysis outputs and tables.

joshzyj/open-scholar-skill

Agent Claude Code

A verification agent that performs Stage 2 verification — comparing the tables and figures embedded/referenced in the manuscript against the statistical claims made in the prose text. Detects misquoted numbers, wrong table references, significance misstatements, directional errors, hypothesis adjudication errors, and…

joshzyj/open-scholar-skill

Agent Claude Code

A verification agent that performs Stage 1 numeric verification — comparing raw analysis outputs (CSVs, HTML tables, R/Python console output) against the formatted tables presented in the manuscript. Detects transcription errors, rounding mistakes, dropped rows/columns, and transformation errors introduced when moving…

joshzyj/open-scholar-skill

Skill Claude CodeCodex

Run data analytics and produce publication-quality tables and visualizations for social science research. Saves regression tables (HTML/TeX/docx), figures (PDF/PNG), an internal analysis log, and a publication-ready Results document (prose + table notes + figure captions). Accepts file paths, inline/pasted data, or…

joshzyj/open-scholar-skill

Skill Claude CodeCodex

Turn unstructured text into validated, structured variables at corpus scale with LLMs: codebook design, dev/gold-set construction, DSPy prompt optimization, a hard reliability gate (Cohen κ ≥ 0.70), scale-out via managed Batch APIs or a local/HPC server, and distillation to a cheap classifier for very large corpora.…