ericluo04

bibcheck

Audit an existing .bib file entry-by-entry against ground truth (DOI, Semantic Scholar, OpenAlex, arXiv) to catch the silent citation errors that pass spell-check but kill peer review — wrong year, mis-cited authors, wrong journal/volume, swapped title-author pairs, fabricated DOIs, and outright hallucinated entries (a real concern with LLM-assisted drafting). Spawns one narrow-focus subagent per entry so each citation gets full attention rather than the gradient-decay that single-pass audits suffer at scale. Use whenever the user says "/bibcheck", "audit my .bib", "verify references", "check my bibliography for errors", "any fake citations", "pre-submission bib audit", "did Claude hallucinate any of these", or before shipping a manuscript / R&R to Marketing Science, JMR, JCR, Management Science. Read-only and audit-only — never adds, edits, or removes entries; produces a PASS/WARN/FAIL report and a `corrected.bib` the user can diff and move into place themselves. Strictly distinct from `/cite` (which ADDS on

blindspot

Shklovsky-inspired peripheral-vision audit of a SINGLE figure or table in a quantitative-marketing paper. Forces a four-quadrant grid — Unexplained Feature, Convenient Absence, Unasked Question, Unexploited Strength — to surface what the author has stopped seeing in a coefficient plot, DiD event-study, conjoint utility plot, predicted-engagement curve, GAN attribute-axis image grid, SAE feature heatmap, or livestream time-series. Use when the user says "/blindspot", "audit this figure", "blindspot check on Figure 3", "what am I missing in this plot", "what's a hostile referee going to ask about this figure", "peripheral vision on Table 4", or "make the stone stony again". Operates at the figure / table level only — manuscript-level adversarial review is `/seven-pass-review`; numeric-claim verification is `/audit-reproducibility`; correctness of the code that built the figure is `/review-paper-code`. They compose — run `/blindspot` early on a fresh figure, then `/seven-pass-review` once the surrounding draft i

compile-latex

Compiles any .tex with auto engine/bib detection, a ranked error report (BLOCKING errors with file:line + suggested \usepackage / package-install fixes, then undefined refs/cites, then threshold-gated overfull/underfull boxes), diff-vs-last-compile, and a conflicted-copy guard. After a clean build it auto-iterates every detected TikZ/pgfplots figure via /tikz-iterate and splices the refined bodies back into the source. Trigger phrases - "/compile-latex", "compile this", "build my paper", "build my deck", "why won't this compile", "what are the latex errors", "did my edit add warnings". Cross-platform MiKTeX (Windows) / MacTeX (macOS) / TeX Live (Linux). Distinct from /tikz-iterate (single figure) and /slide-excellence (deck content review).

council

Spawn up to 5 independent critic subagents in PARALLEL on a target — a research idea, draft, R&R strategy, grant proposal, experimental design, pre-analysis plan, talk plan, IRB protocol, or skill design — collect their raw outputs, then run a SEPARATE synthesis pass that produces a single prioritized action list without majority-voting. Default critic roster is tuned for quantitative-marketing research (skeptic, pre-mortem, methodologist, academic-editor, harsh-referee); `--chef-skill` swaps in a skill-design roster (works with `/skill-creator`). Use whenever the user says "/council", "council review", "get a council on", "spawn critics on", "parallel critique", "kitchen cabinet review", "panel review", "stress-test this idea / plan / skill", or "what would 5 experts say". Target-agnostic and configurable, distinct from `/seven-pass-review` (which is manuscript-bound and runs 7 fixed lens passes on abstract / intro / methods / results / robustness / prose / citations) and from `/evaluate-idea-marketing` and

posterskill

Generate an HTML conference poster from a paper and project website, printable to PDF

referee2

Adversarial cross-language replication audit of an empirical pipeline — reimplement R code in Python (or Python code in R), compare to 6 decimal places, and file a formal referee report covering code correctness, replication-package readiness, output automation, and econometric specification. Use when the user says "/referee2", "cross-language audit", "second-referee check", "reimplement and verify", "stata-style robustness audit", "fresh-eyes audit", "audit my code before submission", or before sending a paper to Marketing Science, JMR, JCR, or Management Science. Stronger than /audit-reproducibility because it does not trust the original language at all — it rebuilds the analysis from scratch in a second language and uses the orthogonality of bugs across languages to surface errors a single-language audit cannot see. Designed for R / Python empirical projects.

seven-pass-review

Run a seven-pass adversarial review of an academic manuscript by spawning 7 parallel review subagents (abstract, intro, methods, results, robustness, prose, citations) and synthesizing their findings into a prioritized revision checklist with 80 / 90 / 95 quality thresholds. Use when the user says "seven pass review", "seven-pass review my draft", "full adversarial review", "deep review my paper", or before submitting / resubmitting to Marketing Science, JMR, JCR, Management Science, or a comparable econ / finance journal. Higher-cost, higher-coverage than /review-paper or /review-paper-light.

slide-excellence

Multi-agent comprehensive Beamer slide review (visual + pedagogy + proofreading, plus TikZ conditionally) for research talks and MBA teaching decks. Use when user says "full review", "excellence pass", "comprehensive check", "review everything", "pre-seminar review", "slide excellence", or before presenting / shipping a deck. Fanout wrapper that spawns slide-auditor, pedagogy-reviewer, proofreader, and (when TikZ present) tikz-reviewer.

tikz-iterate

Iteratively refine a TikZ diagram by compiling it, rendering to PNG, asking the tikz-reviewer sub-agent to evaluate, and applying fixes until APPROVED (or 5 iterations elapse). Designed for Beamer / standalone TikZ snippets in academic papers and lecture decks. Use when the user says "/tikz-iterate", "polish this tikz", "iterate on this diagram until it looks right", "make this figure not look terrible", "compile and review this tikz", or pastes a TikZ block with "fix it visually". Distinct from /slide-excellence (which reviews a whole deck) and the standalone tikz-reviewer agent (which reviews once); /tikz-iterate is the orchestration loop that drives multiple compile/render/refine rounds. Cross-platform — uses latexmk + pdftoppm from MiKTeX (Windows), MacTeX (macOS), or TeX Live (Linux).

academic-pptx

Use this skill whenever the user wants to create or improve a presentation for an academic context — conference papers, seminar talks, thesis defenses, grant briefings, lab meetings, invited lectures, or any presentation where the audience will evaluate reasoning and evidence. Triggers include: 'conference talk', 'seminar slides', 'thesis defense', 'research presentation', 'academic deck', 'academic presentation'. Also triggers when the user asks to 'make slides' in combination with academic content (e.g., 'make slides for my paper on X', 'create a presentation for my dissertation defense', 'build a deck for my grant proposal'). This skill governs CONTENT and STRUCTURE decisions. For the technical work of creating or editing the .pptx file itself, also read the pptx SKILL.md.

litreview

Runs a structured multi-source literature search across arXiv, Semantic Scholar, and OpenAlex MCPs (with Crossref as fallback), dedupes by DOI / arXiv-ID / fuzzy title, scores each unique paper 1–5 for relevance to the user's query, and returns a ranked summarized list with takeaways framed for a quantitative marketing audience. Use this skill whenever the user asks to "find papers on X", "do a lit review on Y", "what's the literature on Z", "any recent work on W", "find references for my paper's intro", or pastes a topic with phrases like "since 2022" / "open access only" / "n=30". Optionally hands off the top hits (score >=4) to `/cite` for batch Zotero adds. Distinct from `/draft` (which writes prose) and `/cite` (which adds a single paper).

preregister

Draft a structured preregistration document (AsPredicted, OSF, or AEA RCT Registry style) from a study spec or free-form description, annotated with MUST / SHOULD / MAY clarity flags. Use when the user says "preregister", "draft a preregistration", "OSF preregistration", "AsPredicted form", "AEA RCT registry", "PAP", "preanalysis plan", or before launching an online experiment, conjoint, vignette study, eye-tracking session, or any data collection / confirmatory analysis they have not yet seen. Output is a LaTeX (`.tex`) or Markdown file the user uploads to OSF / AsPredicted / AEA themselves — this skill writes structure and prose, it does not submit. Designed for quant-marketing experiments (online vignettes, MTurk / Prolific, conjoint, lab eye-tracking, livestreaming field tests) targeting MKSCI / JMR / JCR / MS.

referee-response

Drafts an R&R response letter in the user's R2R style — sectioned by role (Senior Editor / Associate Editor / Reviewer 1 / Reviewer 2), reviewer quotes in `\textit{...}`, location-pinned changes ("see Section 3", "Footnote 7", "Table 4"), "Done." for trivial fixes, polite pushback with cited authority, abandoned drafts kept in `\begin{comment}` blocks. Reads existing `R2R_*.tex` files in the project for tone match and grounds every location pin against the actual `main*.tex`. Use whenever the user asks to "respond to referees", "draft an R&R response", "write the R2R", "address reviewer comments", "respond to the AE", or pastes reviewer comments from an Outlook export / decision letter. Also triggers on "continue the R2R" (extend a partial draft). Output goes to `<project>/R2R_R<n>.tex`.

academic-slides

Create Beamer-inspired academic HTML presentations from scratch or by converting PowerPoint files. Use when the user wants to build slides for a conference talk, lecture, seminar, or thesis defense. Supports theorem environments, KaTeX equations, algorithm pseudocode, and citations. Helps academics discover their preferred visual theme through Beamer-style previews rather than abstract choices.

cite

Resolve a paper (DOI, arXiv ID, title, or URL) to a Zotero entry, append its BibTeX to the active project's .bib file, and return the citation key ready for \citep{}. Use this whenever the user asks to add a citation, "cite this paper", or pastes a paper identifier.

daily-brief

Build the user's morning task brief — pull open tasks from their Notion Tasks DB, score and rank by priority × urgency × type-balance × energy fit (with wait-factor scoring and type-balance rotation), pick the top 3–7 that fit available time, and push the formatted list to their Telegram bot. Saves the selection to today_brief.json so /capture can resolve "1 done" / "2 push thu" replies. Use this skill whenever the user says /daily-brief, asks for "today's tasks", "what should I work on", "morning brief", "rebuild my todo list", "what's on my plate", or wants their daily plan refreshed. Also use when they give a capacity hint like "I have 4 hours today" — pass --hours 4.

draft

Drafts paper sections (Intro, Lit Review, Method, Empirical Setting, Empirics, Discussion, Limitations, or arbitrary paragraphs) in the user's voice — hedged, modest, natbib-apa citations, \Cref cross-references, \emph over \textit, figure/table notes minipages, three-stream lit-review structure. Use this skill whenever the user asks to "draft", "write", "rewrite in my voice", "give me a paragraph on", or "draft a section" for any of their papers. Also triggers on phrases like "write the limitations", "draft an intro for X", "write up the empirical setting", "rewrite this in my voice". Output always goes to a dated draft file under `<project>/drafts/` — never directly into the main `.tex`.

evaluate-idea-marketing

Run an 8-step iterative pre-execution evaluation of a quantitative marketing research idea targeting the Marketing Science / JMR / JCR / Management Science (Marketing) quality bar. Use whenever the user wants to "evaluate a research idea", "score this idea", "is this idea publishable in MKSCI/JMR/JCR/MS", "pivot my idea", "stress-test my project", "find threats to my idea", or pastes a research-question/hypothesis and asks whether it is worth pursuing. Covers idea evaluation, fairness review, pivoting, lit-review threat search with citation verification, and a final verdict — looping on the pivot until the idea scores at least 7/10 against the top marketing-journal bar.

evaluate-idea-science

Run an 8-step iterative pre-execution evaluation of a research idea targeting the broad general-science quality bar — PNAS, Nature Human Behaviour, Nature, Science. Use whenever the user wants to evaluate an idea for "a general-science venue", "PNAS", "Nature Human Behaviour", "Science", "Nature", or asks "is this big enough for a broad audience", "does this have general-interest impact", "could this go to Nature?" Also use when the user has a marketing/economics/CS idea and wonders whether to reframe it for a generalist editor. Covers phenomenon novelty, mechanism strength, generalizability, society-level implications, narrative tightness for a non-specialist editor, lit-review threat search, and a final verdict — looping on the pivot until the idea scores at least 7/10 against the broad-science bar.

replication-package

Bundle a journal-ready replication package (code + data + output + environment + manifest) for a quantitative-marketing paper, zipped to the layout that Marketing Science, JMR, JCR, and Management Science expect at acceptance or revision. Use this skill WHENEVER the user says "/replication-package", "build a replication package", "prepare submission package", "MKSCI replication", "JMR replication", "JCR replication", "Management Science replication", "pre-submission package", "package for replication", "bundle the replication files", "zip up the code and data for submission", "ship the replication archive", or otherwise asks to gather code, data, tables, figures, and environment info into a single submission-ready archive. Also trigger when they say things like "the editor wants the replication files", "ProjectX is going up for replication review", or paste an MKSCI/JMR data-availability email and ask what to send. This skill is the right move even when they do not explicitly say the word "skill" or "package"

review-grant

Run a 6-agent pre-submission panel review for a grant proposal targeting a specified funder or program (NSF, NIH, ERC, HorizonEurope, foundation, etc.). Covers clarity/compliance, internal consistency/deliverables, significance/innovation, research design/feasibility, budget/timeline/team, and an adversarial panel review with funding recommendation.

review-pap

Run a 6-agent pre-submission review of a pre-analysis plan (PAP) for a specified registration target or journal. Covers clarity/pre-specification, hypothesis-outcome consistency, identification, statistical analysis plan, data/operational feasibility, and an adversarial registry/journal review. Supports marketing (MKSCI, JMR, JCR, MS) and economics journal targets.

review-paper-light

Run a fast 2-agent pre-submission check for an empirical paper — focuses on contribution, identification, and causal overclaiming. Designed for marketing (MKSCI, JMR, JCR, MS) and economics/finance papers. Completes in ~1 minute.

review-paper

Run a 6-agent pre-submission referee report for an academic paper targeting a specified journal. Covers spelling/style, internal consistency, unsupported claims, math/notation, tables/figures, and an adversarial top-journal referee evaluation. Supports marketing (MKSCI, JMR, JCR, MS) and economics/finance journals.

skill-creator

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

task-pulse

Answer ad-hoc questions about the user's Notion Tasks DB. Query by project, type, due date, status, or staleness and return a concise readout. Use whenever the user asks "what's open for X", "what's due Friday", "anything stale", "what did I finish this week", "any P0 tasks", or similar status checks against the unified Tasks DB. Read-only — never writes, never marks done, never reschedules.