Research

Runs a deliberative two-model committee with GPT-6 Astra and Claude Opus 5 under a selectable chair. Fable 5.1 is the default chair. `/model-committee-astra` selects Astra with a Sol member, `/model-committee-opus` selects Opus, and `/model-committee-sol` retains the Sol chair with a Terra member. Use when a consequential decision has several defensible options and model families should independently propose, critique, revise, and cross-rank against a predeclared rubric before convergence. Fi...

Audits a methods section against a 45-item checklist synthesized from the APSA Experimental Section rubric, JARS-Quant, CONSORT, and DA-RT. Covers pre-registration and design documentation, subjects and recruitment, randomization and treatment detail, CONSORT-style sample flow and attrition, statistical analysis with sample-size justification and three-tier results labels, conjoint reporting, four validity types, and open-science infrastructure. Use when the user asks whether methods report e...

Reads model uncertainty from token log-probabilities. Covers collecting logprobs and aggregating multi-token labels, confidence tiers and margins for triage, calibration assessment with ECE, Brier scores, and reliability diagrams, responsible downstream use of confidence, and reproducibility archives. Use when the user asks classifier confidence by item, logprobs, top-k tokens, calibration, ECE, Brier, reliability diagrams, or routing low-confidence cases to human review. This is within-model...

Builds or audits a literature review. Produces an evidence map, closest prior work, source clusters, a gap verdict, and a synthesis plan for the introduction. Use when the user asks whether a contribution is novel, who has studied a question, what the literature establishes, how to organize or restructure a review, or wants a reading list, Zotero export, or paper pile turned into a review. Produces a systematic-review protocol scaffold and directs PRISMA or meta-analysis requests to registrat...

Designs and diagnoses list experiments, or the item count technique. Covers whether indirect measurement is warranted, control-list construction against ceiling and floor effects, double-list and direct-question-pairing designs, difference-in-means and maximum-likelihood estimators through ictreg, no-design-effect and no-liar assumptions with ict.test and ict.hausman.test, common-failure corrections, and power planning for the substantial precision penalty. Use when the user measures a sensit...

Drafts a referee report for a journal editor. Produces a recommendation, claim-and-design summary, three to six decision-driving concerns, additional concerns, and a revision plan through five parallel adversarial finders, Breaker, Butcher, Shredder, Void, Situator, a Blue Team error filter, Chief Reviewer synthesis, Tone Guard legal pass, and optional confidential editor note. Use when the user has been asked to referee a manuscript. Self-audits go to paper-review-lite or presubmit, and auth...

Turns theory into falsifiable, pre-registerable hypotheses. Covers DAGs and backdoor closure, design resolution of the FPCI, SATE versus PATE, counterfactual and directional framing, named theoretical and empirical estimands, justified SESOIs, NHST, interval, equivalence (TOST), and minimum-effect tests, scope conditions, and primary, secondary, and exploratory tiers across multi-experiment designs. Use when the user asks to turn a research question or theory into hypotheses, asks whether a p...

Designs and formats publication-quality figures for social-science manuscripts. Covers chart type, scales, color, legend order, self-contained captions, and script-based reproducibility. Use when the user is making, revising, or critiquing a figure for a paper, including coefficient plots, small multiples, distributions, trends, maps, and networks, or asks about labels, colorblind-safe palettes, legend order, direct labels, or captions. Prefer this for manuscript, working-paper, and journal-s...

Provides end-stage QA for a finished figure and table set. Checks inventory and in-text callouts, producing scripts, cross-references and numbering, text-to-evidence consistency, self-contained captions and statistical notes, accessibility, production quality, and SI and replication-package linkage. Marks values requiring image reading as needing author verification rather than guessing. Use when the user prepares a submission and asks to check figures, tables, captions, cross-references, tab...

Fact-checks manuscript claims against their cited sources. Locates each source’s knowledge-base Markdown file and verifies that every in-text claim is supported. Runs a pre-flight gate requiring a clean per-source Markdown knowledge base, one file per cited source, with all raw PDFs converted. Then runs citation-check and audits claim support, overclaiming, direction, scope, and misattribution.

Read or convert documents in a research workflow, including PDF, Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, and CSV. Fires whenever a document must be read, opened, quoted, summarized, searched, extracted, or added to a source library, including "read this paper", "what does this PDF say", "pull the tables out of this", "summarize the attached report", "add this to my sources", or a bare document path. Chooses direct reading or conversion, selects a converter from document structure, a...

Generate 3-5 conceptually distinct approaches labeled by creativity dimension, Novel, Surprising, Diverse, Conventional, and hold for selection before implementation. The --codex mode runs the brainstorm on GPT-6 Astra via codex exec, then has Codex implement the selected approach. Use when a task has more than one non-obvious solution in creative, architectural, or analytical work, and before committing to an approach. Use --codex when another model family should widen the range.

Designs multi-country survey experiments. Covers case selection with comparative logic, instrument localization and TRAPD translation, stimulus origin and composition, per-country power and error management, and pooled or country-specific analysis with measurement-equivalence checks. Use when the user fields the same experiment in several countries, asks which countries to include and why, how to translate or adapt an instrument while preserving comparability, whether to pool or estimate by c...

Reviews an existing conjoint study for inferential threats and returns prioritized findings across design integrity, estimation, measurement error, external validity and behavioral benchmarking, and interpretation. Covers attributes, restrictions, task count, satisficing, randomization, power, estimands, reference levels, subgroups, clustered errors, multiple testing, and the guardrail against reading an AMCE as majority preference. Use when the user asks whether a conjoint design or analysis...

Designs conjoint and factorial-vignette experiments end to end. Covers attribute architecture and randomization restrictions, effective-N power from closed-form AMCE standard errors, treatment realism, AMCE, marginal-mean, and AMIE estimands, forced-choice and rating designs, PAP tiers for conjoint flexibility, and the regression models and R packages for each. Use when the user is planning a conjoint, drafting or critiquing an attribute table, asking how many respondents or tasks are needed,...

Turns a raw Qualtrics conjoint export into an analysis-ready long-format dataset. Covers export settings and metadata rows, implementation identification, wide-to-long reshaping by hand or with cjoint read.qualtrics and cjdata reshape_conjoint, choice-profile alignment, ratings and secondary DVs, attribute translation and factor order, pilot quality diagnostics, and subgroup merges. Use when the user has a conjoint CSV and asks how to clean, reshape, restructure, or debug it, cannot align cho...

Audits a manuscript’s citation layer, including in-text and reference-list parity, fabricated or nonexistent sources, DOIs resolving to another work, APA 7 or named-journal style, completeness, and whether each source supports its attached claim. Verifies against Crossref, OpenAlex, DataCite, and Semantic Scholar, audits LaTeX from cited keys, and marks unchecked items as NOT CHECKED. Use when the user asks to check citations or references, suspects an AI-invented source, wants a .bib checked...

Consult Fable 5.1 as an independent reviewer at maximum reasoning effort. The calling session is the main model, Opus 5 or Sonnet 5, and Fable is the advisor. Use before committing to an interpretation or substantial writing or analysis, when stuck with recurring errors or a non-converging approach, when changing approach, or when a task seems complete and needs a final check. Fallback when the native advisor tool is unavailable. Read-only advisory work that does not edit files.

OCR scanned or image-only corpora with vision-language models, in three phases. Use `evaluate` to compare candidate OCR systems against stratified human ground truth and pick one on measured CER/WER, `run` to build the production pipeline (model selection, image handling, prompts, architecture, batching, accuracy evaluation, reproducibility), and `clean` to correct raw OCR text with LLM and rule-based passes, quality diagnostics, multilingual handling, and span-level provenance. Not for born-...

Specifies and diagnoses structural topic models for survey and experimental text — choosing among STM, LDA, and BERTopic, preprocessing decisions and their consequences, prevalence and content formulas with spectral initialization and a recorded seed, selecting the topic count across semantic coherence, exclusivity and FREX, held-out likelihood, and residuals rather than one metric, interpretation and validation against representative documents, robustness checks, and DA-RT-compliant reportin...

Designs and validates LLM-based text classification for research data — codebook construction, choice of learning regime, model selection and reproducibility, prompt construction, pilot validation against human coding with agreement statistics (kappa, F1), hybrid human-LLM workflows, and reporting model-coded data. Also carries a resumable batch pipeline with a rule-based baseline for coding large sets of repeated free-text values against a closed codebook — occupation, institution, and regis...

Design and format publication-quality tables. Use for column order, row grouping, notes, statistical precision, accessibility, and reproducibility.

Designs survey instruments — item-specific question wording that avoids acquiescence, double-barreling, and leading language, scale construction (points, labeling, polarity, feeling thermometers, index reliability with Cronbach's alpha and McDonald's omega), flow and ordering with treatment placement and buffer items, pretesting through cognitive interviews and soft launches, respondent burden with attention checks and speeding rules, sensitive questions and social-desirability bias, and trea...

Audit fielded survey response data for registered elements, data quality, bot and AI-automation screening, and sample integrity. Emits an appendix-ready quality report.