deck-studio
Skill0 starsChanged 39 days agoMITView on GitHub
Orchestrator playbook plus a code kit for a web deck a speaker would present from: one Vite app, one URL, keys and clicks, steps inside slides, a presenter window with notes and a timer, a PDF export, built from a catalogue of slide types on a 1920x1080 stage, every claim traced to a facts ledger, and verified by an audit script (overflow, step counts, console, reduced motion, no JS, em dash gate) plus contact sheets you look at. Use when the user says "fă-mi o prezentare", "slide-uri pentru", "deck pentru speech / pitch / workshop", "prezentare web", "deck web", "slides for this talk", or wants a presentation that runs in the browser and exports to PDF. Not for .pptx (use the pptx skill).
- Installs packages
- Commits and pushes
2 other projects use a version of this.
Adopted by
Most recently taken up by
- vibe-coding-era/skillstheir own version, first seen Jul 2, 2026
- YujxZJCN/teaching-skillstheir own version, first seen Jun 10, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- YujxZJCN/teaching-skills★ 34same name
- vibe-coding-era/skills★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- landing-studioSkill
- orchestratorSkill
- pet-studioSkill
- ui-bookSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
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
- Distinctive: line 34
Not for: .pptx deliverables (pptx skill), a page people scroll (landing-studio), a document (docs).
- Distinctive: line 46
Nothing outward by an agent: no deploy, no sending, no push, no env.
- Very specific: line 96
Look at
audit/sheet.pngand opendeck.pdf: Every automated check has passed a blank slide at least once. - Distinctive: line 124
After Phase 0: the brief, ten lines.
≈ “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.
Files
Deck Studio
Goal: one talk, one deck, slides that each make one point, presented from a URL with no surprises, with a
PDF for the organisers. The method: brief, facts, outline, one tone direction, one reference slide, fan
out, audit from pixels, rehearse, ship through the owner. Reference build: the MBV Summit deck
(~/Downloads/SPEECH/slide-uri, 2026-10-08; see gallery/README.md).
Language: this playbook is English; the deck's language follows the request ("fă-mi..." means Romanian
with full diacritics). Set <html lang>.
Files in this skill:
CONTRACT.md: stage, steps, navigation, presenter, print, tokens, slide type shape. Agents read it.kit/runtime/,kit/react/,kit/css/: the deck runtime and its CSS (kit/README.md);kit/slides/<type>/withkit/slides/INDEX.md(the catalogue);kit/app-template/andsync-kit.sh.scripts/audit.cjs,scripts/export.cjs,scripts/gate.cjs,scripts/sheet.cjs(scripts/README.md).references/:brief-template.md,facts-ledger.md,outline-template.md,tone-directions.md,slide-rules.md,gotchas.md,cliches.md,handoff-template.md.gallery/README.md: reference decks with their sheets.
When to use
- A talk, a pitch, a workshop, a demo day, a lecture: something presented live or sent as a link.
- Not for: .pptx deliverables (pptx skill), a page people scroll (landing-studio), a document (docs).
Fixed decisions (write them into brief.md; agents obey them)
- Every fact traces to a source line in
facts.md. INFERRED lines never ship. No invented numbers, logos, quotes, dates. Third-party logos never recoloured. - One point per slide. Title says the point, not the topic ("Un singur blocker: tu", not "Context"). Under 40 words on a slide. If a slide needs more, it is two slides.
- Steps reveal, they never rearrange. Space is reserved; nothing jumps. Max 6 steps per slide.
- No em dashes anywhere. Human voice, no clichés (
references/cliches.md). - The deck works without motion and without JS (stacked readable document) and prints to PDF with every step in.
- The stage is 1920 x 1080; design at that size, the runtime scales it. Font sizes in px.
- Nothing outward by an agent: no deploy, no sending, no push, no env.
Phases
Model routing: Opus for facts, outline, the reference slide, integration and the final LOOK; Sonnet for brand ingest, scaffold, slide fan-out, audit runs and fixes with a clear finding.
| Phase | Who | What | Done means |
| 0 Brief | you | brief.md: audience, room (projector? laptop? sent as link?), length in minutes, the one thing people must remember, the one action at the end, sources, language | the owner read it |
| 1 Facts | Opus | facts.md per references/facts-ledger.md | every line sourced, INFERRED flagged |
| 2 Brand | Sonnet | tokens.css values, logo as SVG, fonts local (@fontsource), one glow or texture if the brand has it | contrast checked on the stage (body 7:1) |
| 3 Outline | Opus | outline.md per references/outline-template.md: slide list with type, point, F-ids, steps, notes; about 1 slide per minute, under 1.5; three tone directions per references/tone-directions.md (rhythm, density, imagery), not three palettes | the owner picks a direction and edits the outline |
| ✋ | owner | picks the direction, edits points and order | written into outline.md |
| 4 Scaffold | Sonnet | commands below; src/deck.tsx from the outline with every slide as a placeholder of its type and its notes | npm run build clean, the deck navigates end to end with placeholders |
| 5 Build | Opus ref + Sonnet fan-out | the hardest slide first (usually the demo or a zoom), then one agent per group of slides of the same type, each in its own clone and port; new slide types go back into the skill | each slide passes references/slide-rules.md at its final step and at step 0 |
| 6 Audit | you + Sonnet fixes | scripts/audit.cjs, scripts/export.cjs, scripts/gate.cjs; look at audit/sheet.png and deck.pdf | audit PASS, pdf opened and read, no slide over 40 words unless the owner said so |
| 7 Rehearse | owner, you watch | the owner clicks through once from the real URL in the real browser with the presenter window open; you note every hesitation | the list of fixes is empty or applied |
| 8 Ship | you | commit per wave with a pathspec, handoff from references/handoff-template.md | owner-only commands handed over |
Phase 4: scaffold commands
SK=~/.claude/skills/deck-studio; APP=<repo>/<slug> # brief.md, facts.md, outline.md already in $APP
mkdir -p "$APP" && cp -R "$SK/kit/app-template/." "$APP/" && rm "$APP/sync-kit.sh"
"$SK/kit/app-template/sync-kit.sh" "$APP" --slides title,section,agenda,big-number,cascade-list # any names in kit/slides/INDEX.md
cd "$APP" && npm install && xattr -dr com.apple.quarantine node_modules 2>/dev/null; npm run buildsync-kit.shlays outsrc/kit/{runtime,react},src/styles/{tokens,deck}.css,src/slides/<type>/and a devsrc/preview-deck.tsxmade from everydata.example.ts(one slide per type, in catalogue order), served at/?preview. Re-run after any kit fix; app-owned files (tokens.css,deck.tsx) are written only when missing.- Brand: edit the brand block of
src/styles/tokens.css; put the logo inpublic/logo.svg. npm run devon port 5180 (PORT=5181 npm run dev < /dev/nullto change);npm run buildmakesdist/(single file, opens by double click, works offline).
Phase 5: build
- Reference slide first, Opus, to the full
references/slide-rules.md. Its file is the example in every brief. - Fan-out: one agent per group of slides of one type; each in its own clone (
cp -Rc "$APP" "$TMP/<group>", never a symlink) on its own port; off limits:src/kit,src/styles,src/deck.tsx, other groups. Each slide is checked alone at/?preview#<n>and inside the deck at/#<n>. - The integrator writes
src/deck.tsx(order, transitions, notes) and resolves seams: background handoffs between consecutive slides, a section divider before every chapter, the closing slide's QR and URL.
Phase 6: audit
SK=~/.claude/skills/deck-studio; K=$SK/scripts
npm run build && npx vite preview --port 5190 < /dev/null &
node $K/audit.cjs --url http://localhost:5190/ # every slide, every step, 1920x1080 and 1280x720, rm, no-JS
node $K/export.cjs --url http://localhost:5190/ # deck.pdf + slides/<nn>.png
node $K/gate.cjs src --facts facts.mdLook at audit/sheet.png and open deck.pdf. Every automated check has passed a blank slide at least once.
Then the category audit: would this deck stand next to the three best talks in its field; name the gap or say none.
Phase 7: rehearse
The owner presents once, for real, in the browser they will use, with P open on a second screen. You take notes:
a slide that needs explaining is a slide that failed. Fix, re-audit, re-export.
Owner-only (hand over the exact commands)
- Deploy (
vercel deploy, static hosting), domains, the QR printed on anything. - Sending the PDF or the link to organisers or attendees.
git push, unless asked for in this session.
Sub-agent brief skeleton
Goal: <slide group, the point of each slide, why it matters to the talk>
Scope: you own src/slides/<type>/ and the slides <ids> in your clone <path>, port <51xx>.
Off limits: src/kit, src/styles, src/deck.tsx, every other slide. Never git, deploy or send.
Known facts: CONTRACT.md; outline rows <n..m>; copy slots -> F-ids (quote facts.md exactly); tone direction <x>;
reference slide <path>; three lines from gotchas.md that bite this type.
Done means: references/slide-rules.md, every item, at step 0 and at the final step, 1920x1080 and 1280x720;
you looked at your sheet; server stopped.
Report in under 150 words: files, rule results (n/n), sheet path, open issues. No code.Done means
A speaker could walk on stage with it; the audit is green; the PDF reads on its own; you looked at every slide.
Talking to the owner
- After Phase 0: the brief, ten lines. 2. ✋ After Phase 3: three tone directions and the outline, your pick.
- ✋ After Phase 6: the URL, the sheet, the PDF, the audit summary. 4. Rehearsal notes. 5. Handoff.
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
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.

