agentleFS
liuhlabliuhlab/liulab-skills/skills/lab-presentation/SKILL.md

lab-presentation

Skill1 starsChanged 17 days agoMITView on GitHub

Slides, a slide deck, a presentation, a talk, a pptx or PowerPoint file: use to build, edit or check a deck. The workflow is an audience brief, a text outline, a slide plan page the user approves, the slides, a check of every click state, and speaker notes when wanted. Holds the numbered slide rules, the practice behind them and the `lab-deck` build.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every deck gets steps 1, 2 and 5.

    line 15
  2. Distinctive:

    Read the brief when the deck folder has one.

    line 49
  3. Distinctive:

    Write the outline as text, before any slide exists.

    line 59
  4. Distinctive:

    Write at the brief's level.

    line 64
  5. Very specific:

    Run lab-deck render after each build: one picture per click state.

    line 89
  6. Distinctive:

    Before you report the deck finished, look at every picture once.

    line 91
  7. Distinctive:

    Write notes only when the user asks, or for a formal talk.

    line 97

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

lab-presentation

Build a deck the audience can follow at speaking speed. GLOSSARY.md defines deck, deck folder, audience brief, slide plan, slide rule, deck lint and click state.

Every deck gets steps 1, 2 and 5. Only a deck of one or two slides, on material the audience has seen, skips step 3. Rendered options and speaker notes are extra, each with its own condition below. A recorded talk follows hq-mode's talk playbook after this workflow.

Reference pages

Open a page in full at the step that needs it, not before.

PageOpen it
references/slide-rules.mdBefore the outline, and before any edit to a slide
references/practice.mdWhen you plan the outline or the notes, or when the user questions a rule
references/plan.mdAt step 3, before you write the slide plan
references/build.mdAt step 4, before you write a slide source or run lab-deck
references/checks.mdAt step 5, when you render and lint the built deck
references/round-trip.mdWhen the user has edited the built deck by hand, before you rebuild it
references/images.mdWhen a slide needs a picture: a real figure panel, or one drawn with Codex

1. The audience brief

The brief is brief.toml in the deck folder. Every later check reads it.

FieldHoldsWhen missing
audienceWho is listeningAsk
knowsWhat they already knowAsk
minutesThe length of the talk, as a numberAsk
languageThe language of the slides, such as "en" or "zh"Ask
known_termsA list of the terms and abbreviations they need no help withAsk
room"meeting room" or "hall""meeting room"
use"presented" or "read alone""presented"
levelThe level the talk reports at: "high-level" (what was built, how the work proceeds and why it matters, with detail kept for the notes) or "technical""high-level"

Read the brief when the deck folder has one. Fill every line the request and the source text already answer. Then ask only for the lines still missing, in one message, and say how many there are. Never ask for the room, the use or the level: take the default and say so. Write no outline while one of the first five lines is empty.

For a new deck, agree the lines first. Then run lab-deck new, which makes the deck folder with every asked-for line empty, and write the agreed lines into its brief.toml.

2. The text outline

Write the outline as text, before any slide exists. For each slide give the title, as the sentence that states its conclusion, the body's phrases, and the figure it shows. Start from about one slide per minute and one idea per slide. Settle the content with the user here. Layout comes later.

Write at the brief's level. End with the source of every number.

3. The slide plan

Once the user approves the outline's flow, write plan.html as the plan page says and run lab-deck plan-lint. Then stop: the user approves which slides exist, and each slide's title and picture. It is the last stop: build once it is approved. A later change to the slide set, a title or a picture goes back to the plan; a cut in words does not.

4. The slides

Build the deck with lab-deck, in a deck folder, as the build page says. Commit in the deck folder before you touch the deck and after each build, so the user's hand edits are in history first.

For an edit to a deck that exists, slide rules 14 and 15 come first: say the slide and the exact change in one line, then change only that.

Show 2 to 4 rendered options for a slide only when the user asks, or after the user rejects the slide, as "Options for one slide" on the build page says. Otherwise build one.

5. The check

  • The deck lint stops the build on a blocking failure. Fix the slide. Never loosen a check.
  • Fix every warning of lab-deck lint, or explain it in the reply. None passes in silence.
  • Run lab-deck render after each build: one picture per click state. Look only at the pictures it lists as new or changed.
  • Before you report the deck finished, look at every picture once.
  • At the end, ask the user to open the deck in PowerPoint once. A Repair dialog shows only there, never in a picture.

6. Speaker notes

Write notes only when the user asks, or for a formal talk. Write them last, against the final slides. Each note is an outline of the spoken point, in other words than the slide's. A line the notes never say is a candidate to cut from the slide: propose the cut, and leave the slide alone.

Chinese decks

The same workflow, rules and size floor apply. unslop-zh covers the wording on the slides and in the notes.

Reply

The deck path, the plan page, the slides changed, the pictures you looked at, each lint warning as fixed or explained, and the open choices.

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.