Skip to content
Automation
Skill

/talk

Create and audit presentations (Beamer or Quarto RevealJS). Combines talk creation, visual audit, and compilation. Replaces /create-talk, /visual-audit, /compile-latex (for talks).

From plugin
auto-empirical-research-skills
4.9k200 skills

Claude Code

$ npx -y skills add brycewang-stanford/Auto-Empirical-Research-Skills --skill talk --agent claude-code

SKILL.md

talk.SKILL.md
name: talk
description: Create and audit presentations (Beamer or Quarto RevealJS). Combines talk creation, visual audit, and compilation. Replaces /create-talk, /visual-audit, /compile-latex (for talks).
argument-hint: "[mode: create | audit | compile] [format: job-market | seminar | short | lightning] [--quarto] [file path]"
allowed-tools: Read,Grep,Glob,Write,Edit,Task,Bash

Talk

Create, audit, or compile presentations (Beamer or Quarto RevealJS).

**Input:** `$ARGUMENTS` — mode and format/path.

---

Modes

`/talk create [format]` — Create Beamer Talk

`/talk create [format] --quarto` — Create Quarto RevealJS Talk

Generate a presentation from the paper.

**Agents:** Storyteller (creator) → storyteller-critic (reviewer)

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 |

Workflow

**Step 1: Parse Arguments**

  • **Format** (required): `job-market` | `seminar` | `short` | `lightning`
  • **Paper path** (optional): defaults to `paper/main.tex`
  • **Engine**: Beamer (default) or Quarto RevealJS (`--quarto`)
  • If no format specified, ask the user.

**Step 2: Dispatch Storyteller**

Read the paper and extract: research question, identification strategy, main result, secondary results, robustness checks, key figures/tables, institutional background. Design narrative arc for the chosen format. Build the slide file with shared preamble if available.

The Storyteller follows these design principles:

  • **One idea per slide** — never cram two concepts onto one frame
  • **Figures over tables; tables in backup** — audiences absorb figures instantly; regression tables belong in backup slides where referees can inspect them during Q&A
  • **Build tension** — motivation → question → method → findings → implications
  • **Transition slides between major sections** — signal where the talk is going
  • **All claims must appear in the paper** — the paper is the single source of truth; never add results or claims that are not in the manuscript

Compile with XeLaTeX (Beamer) or `quarto render` (Quarto).

Save to `paper/talks/[format]_talk.tex` (Beamer) or `paper/quarto/[format]_talk.qmd` (Quarto).

**Step 3: Dispatch Storyteller-Critic**

After the Storyteller returns, dispatch the storyteller-critic to review across 5 categories:

| Category | What It Checks | |----------|---------------| | **Narrative flow** | Does the story build properly? Is there a clear arc from motivation through results to implications? Are transitions smooth? | | **Visual quality** | Text overflow, font readability (>= 10pt), figure sizing, consistent formatting, overfull hbox warnings | | **Content fidelity** | Every claim traceable to the paper — no orphan results, no unsupported statements | | **Scope for format** | Right amount of content for the duration — not cramming a seminar into a lightning talk, not padding a short talk to seminar length | | **Compilation** | Does it compile cleanly without errors or warnings? |

Score as advisory (non-blocking). Save report to `quality_reports/[format]_talk_review.md`.

**Step 4: Fix Critical Issues**

If the storyteller-critic finds Critical issues (compilation failures, content not in paper): 1. Re-dispatch Storyteller with specific fixes (max 3 rounds per three-strikes rule) 2. Re-run storyteller-critic to verify

**Step 5: Present Results**

Report to the user: 1. Generated file path 2. Slide count and format compliance 3. Storyteller-critic score (advisory, non-blocking) 4. TODO items (missing figures, tables not yet generated)

---

`/talk audit [file]` — Visual Audit

Check existing slides for layout issues.

Run visual quality checks:

  • Text overflow on any slide
  • Font sizes (>= 10pt for projection)
  • Table readability
  • Figure sizing and labels
  • Consistent formatting
  • Overfull hbox warnings

---

`/talk compile [file]` — Compile Talk

3-pass XeLaTeX compilation for Beamer:

cd paper/talks && TEXINPUTS=../preambles:$TEXINPUTS xelatex -interaction=nonstopmode [file]

For Quarto:

cd paper/quarto && quarto render [file]

---

Principles

  • **Paper is authoritative.** Every claim must appear in the paper.
  • **Figures over tables.** Audiences absorb figures instantly. Put regression tables in backup slides for Q&A.
  • **Less is more.** Especially for short and lightning formats — ruthlessly cut.
  • **One idea per slide.** If you need a second point, make a second slide.
  • **Audience calibration.** Job market = demonstrate rigor and command of the literature. Seminar = sell the interesting result. Short = method and key finding. Lightning = sell the idea in one breath.
  • **Advisory scoring.** Talk scores don't block commits.
  • **Worker-critic pairing.** Storyteller creates, storyteller-critic critiques. Never skip the review.
Read more
@brycewang-stanford
Ships withauto-empirical-research-skills

## 🔥 力荐 · StatsPAI — Stata 与 R 的 Python 平替,为 AI Agent 而生 🔁 Stata / R 平替:reghdfe、csdid、rdrobust、synth、fixest、did 等常用命令在 Python 中一站式替代,一个 import statspai as sp 搞定 DID / RD / IV / SCM / DML / matching 🤖 面向 Agent 设计:结构化结果对象 + 内置 MCP Server,Claude / Codex

4.9k200 skills
Get the whole plugin

Stars on GitHub

Momentum

stars in total
4.9k
Star history of brycewang-stanford/Auto-Empirical-Research-Skills on GitHub, a chart by star-history.com

More skills from other plugins.

agent-builder

agent-builder

Load immediately after an Agent intent, before the first agent_builder_* call. Governs how you create, configure, test, and…

n8n2w209k
@n8n-io@n8n-ioInstall Skill
browser-use

browser-use

Direct browser control via CDP for web interaction: automation, scraping, testing, screenshots, and site/app work. Also covers…

@browser-use@browser-useInstall Skill
agent-browser

agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling…

@vercel-labs-2@vercel-labs-2Install Skill
add-anydoc

add-anydoc

Add local office-document-to-Markdown conversion to NanoClaw agent containers with the pinned Firecrawl AnyDoc CLI. Use when…

@nanocoai@nanocoaiInstall Skill

Questions about talk.

Which coding agents can use talk?
talk installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.