Quarto

Quarto open-source publishing system for R-based reproducible documents. Covers .qmd document format with knitr engine, YAML frontmatter configuration, R code chunks with #| execution options, inline R expressions, rendering to HTML/PDF/Word via quarto render, figure and table output from ggplot2/gt/kable, and cross-referencing. In DAAF Stage 9, its role narrows to literal non-evaluating script-and-log archival, with only bounded existing-data or existing-figure display allowed; broader report-authoring capabilities apply outside Stage 9. Use when the execution language is R and you need to assemble research notebooks, render analysis reports, or produce reproducible documents. Python equivalent: marimo skill.

What is Quarto?

Quarto is an open-source scientific and technical publishing system built on Pandoc:

• Language-agnostic: Supports R (via knitr), Python, Julia, Observable JS
• Multiple formats: Render to HTML, PDF, Word, slides, websites, books
• .qmd format: Plain-text Markdown with YAML frontmatter and executable code chunks
• Git-friendly: .qmd files are plain text -- no binary notebook formats
• CLI-driven: quarto render produces output from the command line

In DAAF, Quarto serves the same role as marimo in Python pipelines: the Stage 9 notebook format that compiles executed scripts into a reviewable document artifact.

Version Notes

This skill targets Quarto 1.7.29 with knitr 1.51 and rmarkdown 2.31.

Key features in Quarto 1.7.x:

• #| (hash-pipe) chunk option syntax is the standard (replaces legacy {r, ...} header syntax)
• Native cross-referencing with @fig-, @tbl-, @sec- prefixes
• YAML frontmatter for all format configuration
• quarto render CLI replaces rmarkdown::render() as the standard rendering path
• Typst output format available (alternative to LaTeX for PDF)

The quarto R package is NOT required. Quarto CLI invokes knitr directly -- the R package is a thin convenience wrapper that adds no functionality beyond what the CLI provides.

How to Use This Skill

Reference File Structure

Each topic in ./references/ contains focused documentation:

File	Purpose	When to Read
quickstart.md	Basic .qmd structure, YAML, R chunks, rendering	Starting a new .qmd document
chunks.md	Code chunk options (`#	` syntax), all execution options
format.md	YAML frontmatter for HTML, PDF, Word, themes, TOC	Choosing and configuring output format
figures.md	Figure output from ggplot2, captions, sizing, cross-refs	Adding figures to documents
tables.md	Table output from gt, kable, kableExtra, cross-refs	Adding tables to documents
daaf-notebook.md	DAAF Stage 9 notebook assembly pattern for .qmd	Required for Stage 9 notebook assembly
rendering.md	quarto render CLI, format selection, parameters, batch	Rendering documents from command line
gotchas.md	Common pitfalls: chunk syntax, YAML, figure paths, cache	Debugging rendering issues

Reading Order

1. New to Quarto? Start with quickstart.md then chunks.md
2. Configuring output? Read format.md then rendering.md
3. Adding figures/tables? Read figures.md or tables.md
4. DAAF Stage 9 notebook? Read daaf-notebook.md (required)
5. Having issues? Check gotchas.md first

Related Skills

Always Load Together (when execution language is R):

• ggplot2 -- Visualization library used for figure chunks in .qmd documents
• data-scientist -- Methodology and validation principles that inform notebook structure

Load for Specific Features:

• gt -- Publication-quality tables in .qmd documents
• fixest -- Fixed effects regression output in .qmd documents

Python Equivalent:

• marimo -- DAAF's Python notebook format (same Stage 9 compilation role)

CRITICAL: Stage 9 is Script COMPILATION, Not Report Authoring

For DAAF research workflows, the Stage 9 Quarto notebook has ONE job: compile executed script contents into .qmd code chunks verbatim.

What Stage 9 IS

• A script viewer -- Paste script code into code chunks
• An audit tool -- Include execution logs to prove what ran
• A file compiler -- Read script files, copy contents, format as .qmd

What Stage 9 is NOT

• NOT a place for new analysis code
• NOT a place for new data transformations
• NOT a place for interactive widgets or Shiny elements
• NOT a report authoring environment (the Report is a separate Stage 11 deliverable)

Stage 9 Notebook Assembly

Stage 9 is handled by the notebook-assembler agent (see .claude/agents/notebook-assembler.md), which:

1. READS .R scripts from scripts/stage{5,6,7,8}_*/
2. EMITS the canonical rich YAML with global execute: eval: false, then one literal R archive chunk per script with exactly one #| code-fold: false, exactly one #| eval: false, and the exact # --- VERBATIM COPY of scripts/<stage>/<filename>.R --- marker as the first nonblank non-option line
3. COPIES the script code beneath that marker literally and un-commented
4. REQUIRES and COPIES each real, non-empty, non-placeholder execution log VERBATIM into exactly one immediately adjacent ::: {.callout-note collapse="true" title="Execution Log"} block with one plain fenced body; a missing or ambiguous log blocks assembly
5. OPTIONALLY ADDS, only after the complete archive/log pair, either a bounded Parquet preview or display of an already-created figure; neither path performs analysis or creates/modifies a figure, and display content is not decompiled

The assembler emits only this canonical form. Bounded legacy containers accepted by scripts/decompile_notebook.R exist solely for intake compatibility and are not valid templates for new Stage 9 output. Structural recognition intentionally centers on the exact VERBATIM COPY marker; the Python/Marimo Plan-template path instead centers on literal mo.md() headers.

ABSOLUTE PROHIBITIONS for Stage 9

Prohibited Element	Why
New library() calls beyond arrow	No new analysis -- copy script code only
New dplyr::filter()	No filtering -- just load and display
New dplyr::mutate()	No transforms -- just load and display
New dplyr::summarize()	No new aggregations -- copy script code only
New ggplot() calls	No new plots -- copy script code only
Shiny runtime elements	Not an interactive app
params: YAML for dynamic reports	Not a parameterized report

The ONLY New Display Content Allowed

Stage 9 permits exactly two bounded display forms:

1. Preview existing Parquet data without transforming it:

df <- arrow::read_parquet("path/to/file.parquet")
dplyr::glimpse(df)
head(df, 20)

2. Display an already-created Stage 8 figure. Prefer non-executing Markdown:

![Figure description](output/figures/existing-figure.png)

When a chunk is needed, it must contain only knitr::include_graphics("output/figures/existing-figure.png") with #| eval: true and #| echo: false.

No filtering, mutation, selection, aggregation, plotting, or figure generation. The preview is bounded to the first 20 rows. (dplyr::glimpse() is namespace-qualified because inspection chunks attach no libraries — a bare glimpse() fails at render time.)

See:

• .claude/agents/notebook-assembler.md for the complete behavioral protocol
• agent_reference/WORKFLOW_PHASE4_ANALYSIS.md Stage 9 for template
• ./references/daaf-notebook.md for the .qmd-specific assembly pattern

---

Quick Decision Trees

"I need to create a .qmd document"

Creating a .qmd?
|- Basic document structure -> ./references/quickstart.md
|- YAML frontmatter options -> ./references/format.md
|- R code chunks -> ./references/chunks.md
|- Rendering to output -> ./references/rendering.md
\- DAAF Stage 9 notebook -> ./references/daaf-notebook.md

"I need to configure output format"

Choosing output format?
|- HTML (default, recommended) -> ./references/format.md
|- PDF (requires LaTeX or Typst) -> ./references/format.md
|- Word (.docx) -> ./references/format.md
|- Multiple formats -> ./references/rendering.md
\- Self-contained HTML -> ./references/format.md

"I need to add figures or tables"

Adding visual elements?
|- ggplot2 figure -> ./references/figures.md
|- Base R plot -> ./references/figures.md
|- Figure sizing/captions -> ./references/figures.md
|- gt table -> ./references/tables.md
|- kable/kableExtra table -> ./references/tables.md
\- Cross-referencing -> ./references/figures.md or ./references/tables.md

"I need to control code chunk behavior"

Chunk execution options?
|- Show/hide code (echo) -> ./references/chunks.md
|- Run/skip code (eval) -> ./references/chunks.md
|- Suppress warnings/messages -> ./references/chunks.md
|- Figure dimensions -> ./references/chunks.md
|- Cache expensive chunks -> ./references/chunks.md
\- Label chunks for cross-ref -> ./references/chunks.md

"Something isn't working"

Having issues?
|- Chunk options not working -> ./references/gotchas.md (#| syntax)
|- YAML parse error -> ./references/gotchas.md (indentation)
|- Figures not showing -> ./references/gotchas.md (figure paths)
|- PDF won't render -> ./references/gotchas.md (LaTeX/Typst)
|- Cache problems -> ./references/gotchas.md (knitr cache)
\- quarto render fails -> ./references/rendering.md + ./references/gotchas.md

File-First Execution and .qmd

Quarto .qmd documents fit DAAF's file-first execution model naturally:

1. WRITE the .qmd file to the appropriate location
2. EXECUTE via CLI: quarto render document.qmd --to html
3. CAPTURE -- output files are written alongside the .qmd (or to --output-dir)

The .qmd source file is the permanent record. The rendered output (HTML/PDF) is the deliverable. Both are stored in the project directory.

For Stage 9 notebooks specifically, the .qmd is assembled from existing executed scripts -- no new code execution occurs during assembly.

Quick Reference

Minimal .qmd Document

---
title: "Analysis Title"
author: "Researcher Name"
date: today
format: html
---

## Section Heading

Narrative text with **bold** and *italic*.

```{r}
#| label: load-data
#| message: false

library(arrow)
df <- read_parquet("data/analysis_data.parquet")
head(df)

Results show r nrow(df) observations.

### Essential Chunk Options

| Option | Values | Purpose |
|--------|--------|---------|
| `echo` | `true`/`false` | Show/hide source code |
| `eval` | `true`/`false` | Execute/skip the chunk |
| `warning` | `true`/`false` | Show/suppress warnings |
| `message` | `true`/`false` | Show/suppress messages |
| `include` | `true`/`false` | Include chunk in output at all |
| `label` | string | Name the chunk (for cross-refs) |
| `fig-cap` | string | Figure caption |
| `fig-width` | number | Figure width in inches |
| `fig-height` | number | Figure height in inches |
| `tbl-cap` | string | Table caption |

### Essential CLI Commands

| Command | Purpose |
|---------|---------|
| `quarto render doc.qmd` | Render to default format |
| `quarto render doc.qmd --to html` | Render to HTML |
| `quarto render doc.qmd --to pdf` | Render to PDF |
| `quarto render doc.qmd --to docx` | Render to Word |
| `quarto preview doc.qmd` | Live preview with auto-reload |
| `quarto check knitr` | Verify R/knitr installation |

## Topic Index

| Topic | Reference File |
|-------|---------------|
| Document structure | `./references/quickstart.md` |
| YAML frontmatter | `./references/quickstart.md`, `./references/format.md` |
| R code chunks | `./references/quickstart.md`, `./references/chunks.md` |
| Chunk options (`#|` syntax) | `./references/chunks.md` |
| HTML format | `./references/format.md` |
| PDF format | `./references/format.md` |
| Word format | `./references/format.md` |
| Themes and styling | `./references/format.md` |
| Figure output | `./references/figures.md` |
| Figure cross-references | `./references/figures.md` |
| Table output | `./references/tables.md` |
| Table cross-references | `./references/tables.md` |
| DAAF Stage 9 notebook | `./references/daaf-notebook.md` |
| Script compilation pattern | `./references/daaf-notebook.md` |
| `quarto render` CLI | `./references/rendering.md` |
| Batch rendering | `./references/rendering.md` |
| Parameterized rendering | `./references/rendering.md` |
| Chunk syntax pitfalls | `./references/gotchas.md` |
| YAML indentation errors | `./references/gotchas.md` |
| Figure path issues | `./references/gotchas.md` |
| knitr cache problems | `./references/gotchas.md` |
| PDF rendering issues | `./references/gotchas.md` |

## Citation

When Quarto is used as the document format, include in the report's
Software & Tools references:

> Allaire, J.J., Teague, C., Scheidegger, C., Xie, Y., & Dervieux, C. (2024). Quarto (Version 1.7) [Computer software]. https://quarto.org

**Cite when:** The analysis notebook or report is delivered as a Quarto document (typically always true in DAAF R pipelines).
**Do not cite when:** Quarto is not used for document delivery.