Plugin

OCR a scanned or image-only corpus with vision-language models, in three phases — evaluate compares candidate OCR systems against stratified human ground truth and picks one on measured CER/WER; run builds the production pipeline (model selection, image handling, prompts, architecture, batching, accuracy evaluation, reproducibility); clean corrects the raw OCR text with LLM and rule-based passes, quality diagnostics, multilingual handling, and span-level provenance. Use when the question is which OCR or VLM to run on a corpus or how accurate one is on your own pages, when scanned documents have to be transcribed at scale, or when raw OCR output needs correction, QA, or a provenance log. Not for born-digital documents with a text layer — those go to doc-to-markdown.

Typeset a working paper or journal submission in house-style LaTeX from any draft — Markdown, Word (.docx), TeX, ODT, RTF, or HTML. Convert with pandoc, wrap in an EB Garamond template, build the PDF with latexmk, and prepare for a specific journal (spacing, page limit, anonymization, disclosures, citation style). Use for "format/typeset/convert my paper to LaTeX", "make a working paper", "prepare this for submission to <journal>".

Launcher and setup wizard for the standalone presubmit CLI, an API-driven adversarial peer-review pipeline of 30-plus stages (Red Team finders, Blue Team defence, verification cascade, legal pass, copyedit) that writes one consolidated review report to disk. Verifies the install and the API key, settles where output lands, picks smoke, standard, or custom mode, launches the run, and reports where the report ended up. Use when the user asks to run presubmit, wants a deep unattended audit of a draft before submission, or needs a math or replication-code audit. The everyday in-session self-audit is paper-review-lite, and refereeing someone else's manuscript is journal-review.

Interviews a researcher, in rounds, until a research idea, design, or draft has no silently assumed decision left — every question numbered, each with a plain-language "why this matters" and a recommended answer, facts fetched by the assistant rather than asked, and every settled decision written to a file. Three stages, idea (a topic or hunch → a falsifiable question and a contribution claim), design (a question → estimand, identification, sample and power, measurement, pre-registration, analysis plan, venue), and defend (a finished design or draft → the objections a reviewer would raise, before the reviewer does). Use when the user says "grill me", "grill this", "stress-test my idea/design/plan", "interview me about this project", "poke holes in this", "what am I assuming", or brings a research idea that is not yet a design. Suitable for BA, MA, and PhD students as well as grant designs; it never answers the research question for the researcher. Adapted for research from Matt Pocock's grill-me. For a softwa

Builds or audits a literature review — evidence map, closest prior work, source clusters, gap verdict, and a synthesis plan that feeds the introduction. Use when the user asks whether a contribution is novel, who else has studied a question, what the literature actually establishes, how to organize or restructure a review section, or wants a reading list, Zotero export, or pile of papers turned into a review. Produces a systematic-review protocol scaffold and hands off to registration (PROSPERO, OSF) and screening tools when the user asks for PRISMA or a meta-analysis.

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