visual-presentation
/visual-presentation
Przygotowuje slajdy i animowane strony HTML do prezentacji: układ, przejścia, animowane wykresy i odsłanianie treści krok po kroku.
- Autor
- Lmandrake ↗
- Repozytorium
- MandrakeWorkflows ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 11.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 3
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 11.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Przygotowuje slajdy i animowane strony HTML do prezentacji: układ, przejścia, animowane wykresy i odsłanianie treści krok po kroku.
Wyzwalacze z opisu
Oryginalny opis (EN)
Front door for slide decks and animated HTML pages made to present or communicate: slides in PowerPoint, Keynote, Google Slides, markdown outlines or HTML, builds and transitions, Morph or Magic Move, animated charts for a talk, scrollytelling and stepped explainer pages. Use whenever the user says "make slides", "draft the deck in markdown", "fix my titles", calls a deck cryptic or too mathy, asks whether a transition or build should be animated, wants a chart animated for a talk, or is building a scroll-driven story page. Decides first whether motion earns its place (default: static), then a
Jak działa
- Purpose
- When to use
- The static-slide core
- The motion decision (make it before any animation is designed)
- How to animate (when the decision is "animated")
- Platform notes
- Quality gate
- Evidence, honestly
- Hand-offs
- Self-check before showing a deck or page
Podgląd SKILL.md
--- name: visual-presentation description: > Front door for slide decks and animated HTML pages made to present or communicate: slides in PowerPoint, Keynote, Google Slides, markdown outlines or HTML, builds and transitions, Morph or Magic Move, animated charts for a talk, scrollytelling and stepped explainer pages. Use whenever the user says "make slides", "draft the deck in markdown", "fix my titles", calls a deck cryptic or too mathy, asks whether a transition or build should be animated, wants a chart animated for a talk, or is building a scroll-driven story page. Decides first whether motion earns its place (default: static), then applies the claim-title slide rules and the motion.* principles. --- # Visual Presentation ## Purpose A deck or a stepped web page is an argument delivered over time. This skill owns two things: the **static-slide core** (every slide makes a claim and defends it) and the **motion decision** (whether anything should move, and if so how). The motion evidence is mostly negative, so the skill's core is knowing when not to animate. ## When to use - Building, outlining, rewriting or critiquing slides, titles, subtitles or speaker notes. - Any question of builds, transitions, Morph, Magic Move, or "should this animate". - An animated chart for a talk, a data video, a scrollytelling or stepped explainer page. - `visual-intent` handed off with a medium of `deck_projected`, `deck_read_ahead`, or an animated or scroll-driven web page. **Not for:** a static figure for a paper or report, a dashboard, a poster or wall display, data art (route to `visual-intent`, `visual-beautiful`, or `scientific-visualization`). A chart that sits on a slide still gets its encoding from `visual-encoding`; this skill decides how the slide and its steps carry it. ## The static-slide core Read `~/dev/MandrakeWorkflows/skills/visual-presentation/references/slides.md` for every deck. Its rules came over unchanged from technical-presentation: title = the takeaway, subtitle = what the slide holds; primary bullets are spoken, secondary text is for the eye; no arithmetic in titles or bold; open with where we stand, then the plan; examples, plain words, no colleagues' names; draft in markdown first when speed is asked for; and its self-check before showing a deck. ## The motion decision (make it before any animation is designed) **1. Default: static.** A static view or a stepped static sequence (hard cuts between steps) is the default (motion.01). A sequence of views is not by itself a reason to move anything. **2. The owner's rule (motion.28).** Animate, or reveal progressively, only when **the change itself is the message** and **its moment lands with the narration** (or the caption, when there is no speaker). If the final frame already carries all the meaning, a build that withholds parts of it has no point per step: it conceals rather than communicates. In the owner's blind viewing, a ranking reshuffle animated onto a slope chart was "just fluff", because the slope lines already showed which item went where; a timed reveal of a sharp drop, landing as the narration named it, was the piece he valued. **3. The jobs motion does well.** Name the job in one sentence, and record why the static alternative does it worse (motion.01 checks). The named jobs: - **Keep marks identifiable across a change of view** of the same data: filter, sort, rescale, regroup, step in time (motion.02), unless the end frame already links start to end (motion.04). - **Draw attention to one change**, one target at a time (motion.15), stopping on the claim (motion.22). - **Show a process that is itself change** in time or space: a physical process, a procedure, a spatial phenomenon (motion.24), with a key-frame static beside it (motion.16). - **Morph between chart types** only when marks map one to one (motion.29). - **A build** only when each step carries its own stated point (motion.23 under motion.28). **4. When not to animate, even with a job.** - Many trajectories, or any reading that compares several frames: small multiples or traces (motion.03). Before-and-after comparison: keep both on screen (motion.04). - Nearly everything would move at once and tracking individuals is not the task: cut (motion.04). - Decoration with no data or argument (entrance flourishes, bounces, spins, animated titles, parallax): cut it (motion.14). - The slide app cannot match the objects across slides: stepped static sequence (motion.26). - "People liked it" is not evidence it communicates (motion.20). Write the decision down before building: ``` motion_decision: static | animated job: <one sentence, or "none"> static_rival: <the best static version, and why it loses (or wins)> lands_with: <the narration line or caption the change lands on> final_frame: <does it carry all the meaning? if yes, why move at all?> ``` ## How to animate (when the decision is "animated") The principles live in `~/dev/MandrakeWorkflows/visualize/principles/principles.yaml` (ids `motion.01` to `motion.29`). Read the entries you apply; do not work from this summary alone. Each entry carries `checks`, `exceptions`, `platforms` and a `validation` line saying how far it has been tested. | Concern | Rule | Ids | |---|---|---| | Frame validity | Every frame of a transition is a valid chart; a mark keeps its datum; unchanged values do not move (MUST) | motion.05 | | Motion off | Every held frame, above all the last, is a complete static chart (MUST) | motion.16 | | Scale | Fixed across steps; stage any rescale apart from value change | motion.06 | | Staging | One stage, or two simple ones, per advance | motion.07, motion.21 | | Easing | Slow-in, slow-out on tracked marks; no overshoot on data | motion.08 | | Timing | About 1 s per data transition (0.7 to 2 s, capped near 2 s); colour or opacity 0.3 to 0.7 s; interface 100 to 500 ms | motion.09, motion.27 | | Attention | About four things to follow; move groups together; one target at a time | motion.10, motion.11, motion.15 | | Pacing | Short segments the speaker or reader advances, not one auto-played run | motion.12, motion.19 | | Words | Narrate with the motion; do not caption the narration over moving marks; label the mover | motion.13 | | Endings | Stop on the claim; motion must not imply a cause the data lack | motion.22 | | Web | Honour reduced motion; no flashing over 3 per second; pause control for long autoplay; scroll steps tied to the reader's scroll | motion.17, motion.18, motion.25 | The recorded tensions between these and their working defaults are sections 10 to 17 of `~/dev/MandrakeWorkflows/visualize/principles/conflicts.md`. Two are **open**: build order versus "a reveal is concealment" (section 16; working default: build only when each step carries a point) and motion.05 being uncheckable mid-morph (section 17; working default: sample mid-transition frames by eye and record that you did). ## Platform notes Read `~/dev/MandrakeWorkflows/skills/visual-presentation/references/platforms.md` before planning motion in a slide app. In short: PowerPoint and Keynote can match objects across slides (Morph, Magic Move), but native charts do not Morph; Google Slides has a smaller set of builds and transitions and no matched-object transition was found (unverified); HTML can do anything, so the gates below are what keep it honest. Slide apps have no per-viewer reduced-motion setting, so the deck must make its point with motion off (motion.16). Slide-app facts beyond the vendor pages are unverified until the owner's manual check runs. ## Quality gate **For HTML pieces:** 1. Implement the step contract in `~/dev/MandrakeWorkflows/visualize/presentation/CAPTURE.md`: `window.__steps` (step count) and `window.__next()` (advance one step; resolves when its motion has settled). Reaching the same resting state under reduced motion. 2. Capture: `python3 ~/dev/MandrakeWorkflows/visualize/tools/motion_capture.py PIECE.html RUN_DIR`, with RUN_DIR under agent storage (`~/agent-storage/mandrakevisualize/motion-test/<piece>/`), never a repo. 3. Gate: `python3 ~/dev/MandrakeWorkflows/visualize/tools/motion_gates.py RUN_DIR`. It fails on motion duration out of band, flashing, missing or blank reduced-motion frames, a final frame that differs from the reduced-motion final frame, and console errors. **Always also** hand every slide and every held frame (the `step_NN_rest.png` captures for HTML; a PDF export for a slide deck) to `visual-render-qa` at final size. **What the gates cannot check.** They do not test comprehension: a piece can pass every gate and still say nothing. They do not check that mid-transition frames are valid charts (motion.05 during a morph), so that stays the builder's eye. Duration and flash detection are sampled at about 0.1 to 0.2 s, so finer timing differences are invisible to them. Slide-app builds are not captured at all; check them by exporting to PDF and by watching the deck run. ## Evidence, honestly The motion rules are grounded in research (animated-transition and trend-animation studies, multimedia-learning work, accessibility standards, vendor guidance; tiers A to C in the registry). They are not yet validated here. Two rounds of model-judge testing (14 briefs, static versus unguided versus principled animation) hit the ceiling and could not separate static from animated: model judges inspect frames at leisure, so transience and pacing cannot hurt them (`~/dev/MandrakeWorkflows/visualize/evaluation/runs/2026-10-28-motion/RESULTS_ROUND1.md`, `~/dev/MandrakeWorkflows/visualize/evaluation/runs/2026-10-28-motion/RESULTS_ROUND2.md`). The only real-time viewing is the owner's blind rating of three animations (n=1), which motion.28 encodes. Following the principles cost no accuracy against unguided animation; whether it helps needs human viewers in real time. Say this when asked how sure the advice is. ## Hand-offs - **From `visual-intent`**, when the medium is slides or an animated or scroll-driven page. Take its intent record (takeaway, audience, medium) as given; each slide's title is a takeaway. - **To `visual-encoding`, then `visual-composition`**, for each chart on a slide or step. This skill does not choose channels or colours; it decides how the chart is staged over time. - **To `visual-render-qa`**, last, for every slide and held frame, after the motion gates. ## Self-check before showing a deck or page - The titles, read alone in order, tell the story as conclusions (slides.md self-check). - A `motion_decision` exists for every moving element, with its job and its static rival. - No reveal of a chart whose final frame already says everything (motion.28). - Every held frame and the motion-off version make the point (motion.16). - HTML: the step contract is implemented and the gates pass; mid-morph frames were checked by eye. - Every slide and held frame went through `visual-render-qa`.
Pliki w paczce (3)
- SKILL.md 11 KB
- references/platforms.md 4 KB
- references/slides.md 5 KB
Wersje
W bazie nie ma innych skilli o nazwie visual-presentation.
Migawka bazy z 13.11.2026 · wersja paczki 8096e4beb501 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100frontend-slides
Tworzy prezentacje HTML z animacjami w jednym pliku bez zależności, od zera lub z konwersji PowerPoint, na pitch decki, wystąpienia konferencyjne i slajdy dla inwestorów. - 82slides
Tworzy i zmienia prezentacje HTML w magic-slides: nowe slajdy, układy, motywy, animacje, konwersję z Keynote i prezentowanie. - 100slide-generation
Tworzy slajdy z ustrukturyzowanych treści w formacie PPTX, Markdown lub HTML na potrzeby agentów. - 100presentation-builder
Tworzy prezentacje i slajdy, w tym w Reveal.js, wystąpienia konferencyjne, wykłady i plakaty, także jako JSON prezentacji. - 100slide-deck
Generuje dopracowaną, gotową do prezentacji stronę HTML z pokazem slajdów z tematu, dokumentu lub notatek. - 100slide-deck-from-doc
Zamienia dokument DOCX, PDF lub Markdown w prezentację PowerPoint z doborem układów i hierarchią treści.