create-lecture

name: create-lecture description: Scaffold a new Beamer lecture or research-talk `.tex` from source papers, notes, or a prior deck — with notation consistency against the project preamble and the user's bold-math conventions wired in. Use when user says "create a lecture on X", "new lecture from these papers", "scaffold a Beamer deck", "build me a research talk on Y", "start a seminar deck", "MBA lecture on Z". Supports two modes: research-talk (MKSCI / JMR / JCR / MS seminars) and pedagogical-lecture (MBA Quant Marketing / Marketing Analytics). argument-hint: "[Topic name] [--mode=research|teaching] [--triage]" allowed-tools: ["Read", "Grep", "Glob", "Write", "Edit", "Bash", "Task"] context: fork

Lecture / Talk Creation Workflow

Source. Base skill adapted from pedrohcgs/claude-code-my-workflow. The Beamer-pedagogy workflow phases (intake → paper analysis → structure proposal → draft slides → polish), the notation-consistency check against the project preamble, and the dual research-talk / pedagogical-lecture mode split are Pedro H.C. Sant'Anna's. The opt-in --triage Step-0 gate (audience + closing-claim pre-write check) is adapted from Scott Cunningham's MixtapeTools:beautiful_deck.

Personalization

This skill resolves placeholders against ~/.claude/state/personal_config.json. See _config/README.md and _config/personal_config.example.json for setup. If the config is missing or a needed field is unset, the skill must surface an error to the user and refuse to proceed rather than guess.

Required config fields:

personal_config.paths.overleaf_root — root containing project subdirs.
personal_config.projects[] — each entry should declare overleaf_subdir and optional preamble_tex / bib_file.

Purpose

Create a beautiful, pedagogically clean Beamer deck — either a research talk for a seminar or a teaching lecture for MBAs.

This is a collaborative, iterative process. The user drives the vision; Claude is a thinking partner.

MODE SELECTION

Parse --mode from $ARGUMENTS. If absent, infer from context:

research mode — invited seminar, conference talk, job-talk style, JMR/MKSCI submission talk. Audience: peer faculty + PhD students. Tone: assumes domain priors, dense with results, citation-heavy, single running empirical thread.
teaching mode — MBA Quant Marketing or Marketing Analytics lecture. Audience: MBAs with limited stats background. Tone: motivation-first, every formal definition gets a worked example within 2 slides, embedded Socratic questions, 2-3 fragment reveals per deck.

Both modes share the constraints below. Differences are flagged inline as [research] / [teaching].

Modes

Default (no flag). Mode inferred from path/keywords as described above; scaffold proceeds straight into Phase 0. Behavior unchanged from prior versions of this skill.
--triage (opt-in). Before any .tex is written, run the Pre-Write Triage gate below. Three questions, asked once, then the skill returns to its normal flow. Discipline, not bureaucracy — the gate exists because a deck whose closing claim isn't yet a single sentence will burn an hour of slide-shuffling later.

Pre-Write Triage (only when `--triage` is passed)

Ask the user, in order:

Audience — pick one: peer researchers (econometric / methodological) / MBA students (applied / decision-relevant) / mixed academic-practitioner / brownbag / dissertation committee.
Single-sentence closing claim — "After this talk/lecture, the audience should leave knowing _____." One sentence. Not a bullet list.
Mode confirmation — research-talk vs. pedagogical-lecture. Overrides the inferred mode if the user picks the other one.

If the user cannot articulate the closing claim in one sentence, refuse to scaffold and surface verbatim: "State the closing claim in one sentence first; the deck structure depends on it." Do not proceed to Phase 0 until the sentence exists.

Record the three answers and carry them into the Pre-Flight Report (audience → "Audience & narrative position"; closing claim → "Pedagogical goal"; mode → "Mode").

CONSTRAINTS (Non-Negotiable)

Read the project preamble FIRST — the user's bold-math macros (e.g. \bphi, \bomega, \bz, \bw, \bs), citation style (natbib-apa: \citep, \citealt, \citet), and cross-ref macro (\Cref) all live in preamble.tex.
Every new symbol MUST be checked against the preamble's notation conventions — flag conflicts before drafting.
Motivation before formalism — no exceptions, both modes.
[teaching] Worked example within 2 slides of every definition.
[research] A running empirical application (one of the user's own projects, or the topic's canonical example) threaded throughout.
Max 2 colored boxes per slide.
No \pause or overlay commands unless the user explicitly asks — they hurt PDF export and Beamer→Overleaf diff readability.
Transition slides at major conceptual pivots.
All citations verified against the project's .bib file; use \citep/\citet per natbib-apa.
Work in batches of 5-10 slides — share for feedback, don't bulk-dump.

WORKFLOW

Phase 0: Intake & Context (Pre-Flight Report required)

Read the inputs, then produce a Pre-Flight Report before Phase 1.

Inputs to read:

Project preamble.tex (look under <OVERLEAF_ROOT>/<PROJECT_SUBDIR>/ or a sibling folder).
Project .bib file(s) so citation keys can be verified.
Any source papers / existing slides the user provided.
A prior deck's .tex if this is a follow-on lecture.

Required Pre-Flight Report block:

## Pre-Flight Report

**Mode:** [research / teaching]

**Sources read:**
- [source 1]: [one-line takeaway — what notation, what result, what figure]
- [source 2]: ...

**Preamble check:**
- Preamble path: [path]
- Existing bold-math macros available: [list from preamble]
- New symbols this deck will introduce: [list]
- Conflicts with preamble: [none / specific clashes]

**Citation check:**
- `.bib` path: [path]
- Citation keys to be used: [list]
- Missing from .bib: [none / list — user must add via `/cite` skill]

**Audience & narrative position:**
- [research] Venue + audience priors. What's the one-slide takeaway?
- [teaching] Where does this lecture sit in the course arc? What did the previous lecture end on?

**Pedagogical goal:** [one sentence]

**Running application:** [which real-world example threads through this deck]

State the goal, get the user's confirmation, then proceed.

Fresh-project fallback. If no preamble exists yet (brand-new Overleaf project), propose a minimal starter preamble with the user's standard macros and natbib-apa citation setup. Present for approval, write it, then continue to Phase 1.

Phase 1: Source Analysis (When Papers Provided)

Split each paper into chunks; extract key ideas, results, notation.
Map paper notation → the user's project notation.
Identify slide-worthy content (theorems, key plots, summary tables).
Present summary for approval.

Phase 2: Structure Proposal

Propose outline.
- [research] 5-Act: Motivation → Setting → Identification/Method → Results → Implications.
- [teaching] 3-Part: Why-this-matters → Core idea (build incrementally) → Apply-it (worked example + small exercise).
List TikZ diagrams and figures needed.
List new notation to introduce, mapped to preamble macros.
GATE: User approves before Phase 3.

Phase 3: Draft Slides (Iterative)

Work in batches of 5-10 slides.
Check every new symbol against the preamble; use existing bold-math macros rather than re-defining bolds.
Use \citep{} for parenthetical, \citet{} for in-line, \citealt{} for "see also" lists.
Use \Cref{...} for all internal cross-refs.
Apply mode-specific patterns:
- [teaching] Embed 2-3 Socratic prompts; use 3-5 fragment reveals at problem→solution moments; pair every definition with a worked example.
- [research] Lead each section with a one-sentence claim; keep one running empirical thread; budget ~1 slide per minute of talk.

Phase 4: Figures & Code

R or Python scripts following the user's project conventions (no Stata).
TikZ diagrams inline in the Beamer source so the deck stays self-contained.
Save plot data (RDS / parquet) alongside scripts for reproducibility.

Phase 5: Polish & Compile

Compile via latexmk -pdf <deck>.tex from the Overleaf project root.
Run /slide-excellence for the full multi-agent pass.
If any TikZ blocks exist, expect tikz-reviewer to flag label-overlap issues — iterate until APPROVED.

Post-Creation Checklist

[ ] Deck compiles without errors via latexmk
[ ] No overfull hbox > 10pt
[ ] All citations resolve against the project .bib
[ ] [teaching] Every definition has motivation + worked example
[ ] [research] Running empirical thread is visible across sections
[ ] Max 2 colored boxes per slide
[ ] [teaching] 2-3 Socratic questions embedded
[ ] Transition slides between sections
[ ] At least 1 running application threaded throughout
[ ] New notation does not collide with preamble macros
[ ] /slide-excellence run; critical issues addressed

Failure modes

Preamble not found. Don't invent bold-math macros from scratch — ask the user to point to the project's preamble, or scaffold a fresh one as in Phase 0 fallback.
Citation key invented. Never fabricate a .bib key. If a citation is missing, list it under "Missing from .bib" in the Pre-Flight Report and have the user run /cite first.
Stata / SAS code suggested. Never. This skill targets R + Python only.
Mode mismatch. A research talk with MBA-style "let's think about why this matters" framing will feel patronizing to a JMR audience; an MBA lecture with proof-density will lose the room. If the mode inference is uncertain, ASK before drafting.

Out of scope

Compiling for the user — the user compiles in Overleaf; this skill only suggests the latexmk command for local previews.
Editing PowerPoint or Google Slides — Beamer only.
Generating Quarto-flavored slides.
Reviewing an existing deck (use /slide-excellence for that).