create-talk

name	create-talk
description	Generate academic presentations (Beamer .tex or Quarto RevealJS .qmd) by dispatching the storyteller agent (creator) and discussant agent (critic). Supports 4 formats — job market, seminar, short, lightning. Derives all content from paper/main.tex.
disable-model-invocation	true
argument-hint	[format: job-market | seminar | short | lightning] [output: beamer | quarto (default: beamer)] [paper path (optional)]
allowed-tools	["Read","Grep","Glob","Write","Edit","Task","Bash"]

Create Talk

Generate an academic presentation by dispatching the storyteller (creator) and discussant (critic).

Input: $ARGUMENTS — format name, output type (beamer or quarto), optionally followed by paper path.

---

Workflow

Step 1: Parse Arguments

• Format (required): job-market | seminar | short | lightning
• Output type (optional): beamer (default) | quarto
• Paper path (optional): defaults to paper/main.tex
• If no format specified, ask the user before proceeding.

Format Constraints

Format	Slides	Duration	Content Scope
Job market	40-50	45-60 min	Full story, all results, mechanism, robustness
Seminar	25-35	30-45 min	Motivation, main result, 2 robustness, conclusion
Short	10-15	15 min	Question, method, key result, implication
Lightning	3-5	5 min	Hook, one result, so-what

Output Format Details

Beamer (.tex)

• Output: talks/beamer/[format]_talk.tex
• Compile: latexmk -pdf -cd talks/beamer/[format]_talk.tex
• Figures: \includegraphics[width=0.85\textwidth]{../../paper/figures/fig.pdf}

Quarto RevealJS (.qmd)

• Output: talks/quarto/[format]_talk.qmd
• Compile: quarto render talks/quarto/[format]_talk.qmd
• Figures: ![](../../paper/figures/fig.pdf){width=85%}
• YAML: format: revealjs, theme: simple, embed-resources: true, slide-number: true

Step 2: Launch storyteller Agent

Delegate to the storyteller agent via Task tool:

Prompt: Create a [format] talk from [paper] in [beamer|quarto] format.
Read paper/main.tex fully. Extract: research question, identification strategy,
main result (with exact value and units), secondary results, robustness checks,
key figures/tables (with file paths), institutional background.
Design narrative arc for [format] format.
Build [Beamer .tex | Quarto .qmd] file. Use figures from paper/figures/ directly.
Compile to verify it runs cleanly.
Save to talks/[beamer|quarto]/[format]_talk.[tex|qmd]

The storyteller follows these principles:

• One idea per slide
• Figures > tables (move tables to backup slides)
• Build tension: motivation → question → method → findings → implications
• All claims must appear verbatim in paper/main.tex (single source of truth)
• Notation must match paper exactly (same subscripts, Greek letters, variable names)

Step 3: Launch discussant Agent (Talk Critic)

After storyteller returns, delegate to the discussant agent:

Prompt: Review the talk at talks/[format]/[format]_talk.[tex|qmd].
Also read paper/main.tex to verify content fidelity.
Format: [Beamer | Quarto]
Type: [format]
Run all 8 audit categories. Score as advisory (non-blocking).
Save report to quality_reports/[format]_talk_review.md

Step 4: Fix Critical Issues

If discussant finds Critical issues (compilation failure, content not in paper):

1. Re-dispatch storyteller with the specific issues listed (max 3 rounds per three-strikes.md)
2. Re-run discussant to verify

Step 5: Present Results

1. Generated file path and format
2. Slide count and format compliance
3. discussant score (advisory, non-blocking)
4. Any TODO items (missing figures, unresolved references)

---

Output

• Beamer: talks/beamer/[format]_talk.tex
• Quarto: talks/quarto/[format]_talk.qmd

---

Principles

• Paper is authoritative. Every claim must appear in paper/main.tex.
• Same figures. Never redraw figures — use the exact same files from paper/figures/.
• Less is more. Especially for short and lightning — ruthlessly cut.
• Audience calibration. Job market = rigor. Seminar = interesting result. Lightning = sell the idea.
• Advisory scoring. Talk scores don't block commits.
• Worker-critic pairing. storyteller creates, discussant critiques. Never skip the review.