agentleFS
flonatflonat/flonat-research/skills/beamer-deck/SKILL.md

beamer-deck

Skill150 starsChanged 10 days agoMITView on GitHub

Create an academic presentation as a LaTeX Beamer source and reviewed PDF with an original theme. Use when the requested deliverable is a conference, seminar, or lecture deck in Beamer. Not for PowerPoint or RevealJS; use $pptx or $quarto-deck.

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 10 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create polished, zero-warning Beamer decks for academic contexts: seminars, conference talks, teaching lectures, and working decks for …

    line 15
  2. Very specific:

    Python: Always use uv run python. Never bare python, python3, pip, or pip3.

    line 32
  3. Distinctive:

    One idea per slide: Not a guideline. A law. If a slide has two ideas, split it.

    line 35
  4. Very specific:

    Use the unified template: Always use \usepackage[<institution>]{user-beamer} from templates/beamer/.

    line 36
  5. Distinctive:

    Start at 100, deduct per issue found, apply verdict.

    line 58
  6. Distinctive:

    Read project files, content sources, and audience brief.

    line 81
  7. Very specific:

    Check for existing .bib files in the project.

    line 88
  8. Very specific:

    If using citations: add \addbibresource{references.bib} or \bibliography{} as appropriate

    line 107
  9. Very specific:

    After latex resolves errors, address remaining warnings (which autofix does not fix):

    line 111
  10. Distinctive:

    Circuit breaker — do not spiral: If the same compile error persists after 3 different fix attempts, STOP.

    line 118
  11. Very specific:

    Stop editing the .tex file.

    line 120
  12. Distinctive:

    Ask the user how to proceed.

    line 123
  13. Very specific:

    PDF visual inspection: Run uv run python scripts/pdf-to-images.py <deck>.pdf to convert pages to images, then inspect each image for text …

    line 132
  14. Distinctive:

    Read both reviewer reports

    line 148
  15. Distinctive:

    Verify zero warnings in the log

    line 151
  16. Distinctive:

    If significant changes were made, loop back to Phase 5 for another review round

    line 153
  17. Distinctive:

    Use as inspiration — always create an original palette for each deck.

    line 163

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Beamer Deck Skill

Generate academic Beamer presentations with original themes, rhetoric-driven structure, and multi-agent review. Internalises Scott Cunningham's rhetoric framework and implements an adversarial review workflow.

Purpose

Create polished, zero-warning Beamer decks for academic contexts: seminars, conference talks, teaching lectures, and working decks for coauthors. Every deck gets a custom theme, assertion-driven titles, and parallel review by rhetoric and graphics sub-agents.

NOT for project status updates — use project-deck for those.

When to Use

  • Academic seminar presentations
  • Conference talks (15–45 min)
  • Teaching lectures (undergraduate or PhD)
  • Working decks for coauthors or supervisors
  • Any presentation that needs rhetoric discipline and visual quality

Critical Rules

  1. Build artifacts go to out/, PDF stays in the source directory. Use the latex pre-flight to install the fail-closed canonical .latexmkrc; never hand-write a partial config or copy block. Use latex for compilation and engine resolution.
  2. Python: Always use uv run python. Never bare python, python3, pip, or pip3.
  3. Fix ALL warnings. Overfull hbox, underfull hbox, overfull vbox, underfull vbox — no matter how small. Parse the .log file. Recompile until clean.
  4. Titles are assertions, not labels. "Distance increases abortion rates" — not "Results". Every frame title states a claim.
  5. One idea per slide. Not a guideline. A law. If a slide has two ideas, split it.
  6. Use the unified template. Always use \usepackage[<institution>]{user-beamer} from templates/beamer/. Options: preset-a, preset-b, preset-c, preset-d, plain. Never use default Beamer themes (Warsaw, Madrid, etc.) as-is, and never create one-off preambles. See [templates/beamer/README.md] for available custom commands and how to add institutions.
  7. Code-first figures. Generate figures via R or Python scripts before inserting. Never use placeholder images. Always save the script alongside the figures — never generate a figure without preserving the code that created it.
  8. If a .bib file is used, validate it. Cross-reference all \cite{} keys against the bibliography file. See bib-validate for the full protocol.
  9. Never define parameterized TikZ styles inside a frame. # inside a Beamer frame body is consumed by the frame parser before TikZ sees it, producing Illegal parameter number errors that cascade and resist all downstream fixes. Define ALL \tikzset{...} and .style={..., #1} entries in the preamble; use them inside frames. See ../shared/tikz-rules.md Rule 11.

Rhetoric Principles

Full framework (Three Laws, MB/MC, Aristotelian Triad, Narrative Arc, Pyramid Principle, Devil's Advocate): ../shared/rhetoric-principles.md

Scott's original essay: resources/academics/scott-cunningham/MixtapeTools/presentations/rhetoric_of_decks.md


Quality Scoring

Apply numeric quality scoring using the shared framework and skill-specific rubric:

  • Framework: ../shared/quality-scoring.md — severity tiers, thresholds, verdict rules
  • Rubric: references/quality-rubric.md — issue-to-deduction mappings for this skill

Start at 100, deduct per issue found, apply verdict. Compute the score in Phase 7 and report it in the final output.

Context-Specific Guidance

See ../shared/rhetoric-principles.md for the full Aristotelian Triad framework. Context-specific Ethos/Pathos/Logos balance and adjustments are applied in Phase 2 based on audience type (academic seminar, conference talk, teaching lecture, working deck).


Workflow: 7 Phases

You (orchestrator)
├── Phase 1: Gather context        (direct)
├── Phase 2: Design structure       (direct)
├── Phase 3: Build deck             (direct)
├── Phase 4: Fix all warnings       (direct)
├── Phase 5: Rhetoric review        (sub-agent — Explore)
├── Phase 6: Graphics review        (sub-agent — Explore)  ← parallel with Phase 5
└── Phase 7: Apply & finalise       (direct)

Phase 1: Gather Context (Direct)

Read project files, content sources, and audience brief. Ask the user clarifying questions:

  • Audience: Who is this for? (seminar, conference, teaching, coauthors)
  • Duration: How long is the talk?
  • Content source: Paper draft? Notes? Existing slides? Code output?
  • Special requirements: Specific figures, institutional branding, language?

Check for existing .bib files in the project. If citations are needed, note this for Phase 3.

Phase 2: Design Structure (Direct)

  1. Choose rhetoric balance based on audience (see table above)
  2. Outline slide sequence with assertion titles — write each title as a claim
  3. Plan narrative arc — consult docs/reference/talk-design.md for format-specific arcs (empirical, structural, theory, descriptive) and audience calibration
  4. Choose institution option for user-beamer — colours and fonts are set by the template
  5. Identify figures needed — which need to be generated via code?

Present the outline to the user for approval before building.

Phase 3: Build Deck (Direct)

  1. Generate figures first — run R/Python scripts, save to figures/
  2. Write .tex file using the unified template: \documentclass[aspectratio=169,11pt]{beamer} + \usepackage[university]{user-beamer} (or other institution option). Use \fbinstitute and \fbemail for metadata. Custom commands: \contribcard, \phasecircle, \accentbox, \highlightbox, standoutframe environment.
  3. Use 16:9 aspect ratio (already in the documentclass above)
  4. Install the canonical .latexmkrc through the latex pre-flight if it is missing; classify any divergent existing config before migration
  5. Compile using latex — this handles missing packages, font conflicts, citation key mismatches, and stale cache automatically
  6. If using citations: add \addbibresource{references.bib} or \bibliography{} as appropriate

Phase 4: Fix All Warnings (Direct)

After latex resolves errors, address remaining warnings (which autofix does not fix):

  1. Parse out/*.log for overfull/underfull hbox/vbox warnings
  2. Fix every single one — adjust text, resize figures, tweak \parbox, etc.
  3. Recompile
  4. Repeat until the log is clean

Circuit breaker — do not spiral. If the same compile error persists after 3 different fix attempts, STOP. Do not try a fourth approach. Instead:

  1. Stop editing the .tex file.
  2. Quote the exact error line from the .log.
  3. List the 3 approaches tried and why each failed.
  4. Ask the user how to proceed.

What counts as "the same error": any error that persists at the same line family after a fix (e.g., Illegal parameter number moving from line 568 to 572 is still the same error — that is the diagnostic signature in tikz-rules.md Rule 11). A fresh Overfull on a different slide after fixing the first is a new error — reset the counter.

The cost of stopping is 2 minutes; the cost of spiraling is an hour of edits that progressively obscure the original cause. This rule overrides "recompile until clean" — zero tolerance for warnings does not mean infinite attempts on the same fix.

"Compilation success does not mean visual success." Also check for silent visual errors:

  • TikZ: shape constraints forcing label misplacement, coordinate misalignment. If the deck contains TikZ diagrams, run the 6-pass verification from ../shared/tikz-rules.md — compute Bezier depths, check gaps, verify label fit, check shape boundary clearance.
  • matplotlib/ggplot: axis labels cut off, legend obscuring data, text sizing. If figures use curved arrows (arc3), compute Bézier positions using the helper functions in ../shared/tikz-rules.md § Matplotlib Extension — never guess where curves pass. Check label-to-shape clearance (Boundary Rule) and use anchor-based centering for text pairs.
  • PDF visual inspection: Run uv run python scripts/pdf-to-images.py <deck>.pdf to convert pages to images, then inspect each image for text overflow, element overlap, font readability, and alignment issues that are invisible in the log

Phase 5: Rhetoric Review (Sub-Agent — Explore)

Launch a sub-agent to review the .tex file against 7 criteria: narrative arc, MB/MC balance, title quality, one-idea-per-slide, transitions, Aristotelian balance, and pyramid principle.

Full prompt template: references/review-prompts.md § Rhetoric Review

Phase 6: Graphics Review (Sub-Agent — Explore)

Launch in parallel with Phase 5. Reviews TikZ diagrams, figure sizing, table formatting, colour consistency, typography, and numerical accuracy.

Full prompt template: references/review-prompts.md § Graphics Review

Phase 7: Apply and Finalise (Direct)

  1. Read both reviewer reports
  2. Incorporate feedback — prioritise Critical and Needs Work items
  3. Recompile
  4. Verify zero warnings in the log
  5. If using a .bib file: validate all \cite{} keys resolve correctly (check log for Citation .* undefined). See bib-validate for the full cross-referencing protocol.
  6. If significant changes were made, loop back to Phase 5 for another review round
  7. Compute quality score — read references/quality-rubric.md, log all issues from Phases 4-6, compute score and verdict
  8. Confirm final PDF is in the source directory (copied from out/ by .latexmkrc)

Reference Palettes

Three starting palettes (Professional, Energetic, Academic) in both LaTeX and CSS formats: ../shared/palettes.md

Use as inspiration — always create an original palette for each deck.


Output Checklist

A completed deck directory should contain:

project/
├── deck.tex              # Main Beamer file (uses user-beamer.sty)
├── deck.pdf              # Compiled PDF (copied from out/ by .latexmkrc)
├── .latexmkrc            # Canonical build config
├── out/                  # Build artifacts only
├── figures/              # Generated figures (if any)
│   ├── figure_1.png
│   └── ...
├── scripts/              # R/Python scripts that generated figures (if any)
│   ├── figure_1.R
│   └── ...
└── references.bib        # Bibliography (if citations used)
  • PDF compiles with zero warnings
  • All frame titles are assertions
  • One idea per slide
  • Narrative arc: Problem → Investigation → Resolution
  • Rhetoric review completed (Phase 5)
  • Graphics review completed (Phase 6)
  • If .bib used: all \cite{} keys validated (see bib-validate)
  • Quality score computed and reported

Cross-References

SkillWhen to use instead/alongside
project-deckFor project status updates (supervisor meetings, coauthor handoffs)
latexDefault compiler — used in Phase 3 for error resolution and citation audit
latexFor manual compilation config details, .latexmkrc setup, engine selection
proofreadFor post-hoc review of text quality in the deck
bib-validateFor thorough bibliography cross-referencing when citations are used
literatureFor finding and verifying citations to include
quarto-deckFor HTML presentations (teaching, informal talks) instead of PDF
quarto-courseFor full course websites with multiple lectures, exercises, and navigation

Scott's full rhetoric essay: resources/academics/scott-cunningham/MixtapeTools/presentations/rhetoric_of_decks.md Scott's deck generation prompt: resources/academics/scott-cunningham/MixtapeTools/presentations/create_deck_prompt.md Scott's example decks: resources/academics/scott-cunningham/MixtapeTools/presentations/examples/

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.