chore: 1.1.0 release

skills/open-kimi-ppt/reference/design_system/academic/blue-line-courseware/design.md

@@ -0,0 +1,81 @@
+# Blue-Line Courseware · Academic STYLE DESIGN SYSTEM
+
+## PART A — Category Baseline (General Minimum)
+
+A rigorous, formally typeset academic presentation benchmarked against top-tier research talks and thesis defenses (NeurIPS/CHI/SIGGRAPH main-conference standards), with clarity of argument taking precedence over decoration. The deck must read as a reproducible scientific claim, not as marketing collateral. Use academic language throughout: standard terminology for research questions, hypotheses, evidence, significance, contributions, and limitations, without marketing or business rhetoric. Shift the visual emphasis by discipline: in theoretical work, formulas and proof steps occupy the main visual field; in computer science and engineering, system diagrams, code, execution traces, and performance curves provide the backbone; in experimental research, formulas, charts, and images cluster tightly around a single conclusion; in the humanities and social sciences, archival material, maps, manuscripts, and other textual evidence add narrative depth without compromising academic rigor.
+
+Content and layout:
+1. Use a fixed argumentative spine: research question → gap in related work → data and methods → main results → ablation/analysis → limitations → future work. A short deck may compress this sequence, but must not reorder it.
+2. Build evidence-first slides: results tables must include baseline comparison columns, comparative charts, and architecture/process diagrams; present the core contribution as a formally typeset equation.
+3. Maintain moderate information density: one argument per slide, with evidence organized around that argument. Slides should be composed, not merely filled—white space is a valid means of separating stages of the argument.
+
+Visual style:
+1. Aesthetic: restrained, precise, and scholarly. No promotional copy, decorative photography, or clip-art icons.
+2. Typography: use a clear sans serif (render with Liter / Chinese: MiSans) for body copy and labels; when the source style uses serif typography, titles may also use serif. Equations must be formally typeset (KaTeX quality), with standard equation numbering and italic variables.
+3. Color: use one two-color system per deck—one primary structural color and one accent color, with gray establishing hierarchy; use a white or near-white background; reserve one consistent warning color for negative or ablation signals.
+4. Graphics: use thin, precise vector lines in diagrams; minimal horizontal rules in tables; axis labels, units, and sample definitions in charts; no decorative 3D effects, shadows, or gradients.
+
+Structure and discipline (distilled from deliverables established as benchmarks):
+1. Every content-slide title must state an argument (a finding, not a section label—“Lapses in Attention Can Be Predicted 10 Seconds in Advance,” not “Experimental Results”), followed by evidence that supports it.
+2. Align claims with evidence: every quantitative conclusion on a slide must be supported on that same slide by a chart, table, or cited result; no orphaned numbers.
+3. The main results table must include baseline comparison columns and identify the best value; show relative improvement alongside absolute values.
+4. Method slides must include reproducibility details: dataset size, annotation protocol, inter-annotator agreement, and key hyperparameters. Show only the core idea as an equation; move derivations to appendix slides.
+5. Every figure must be self-contained: include axis labels with units, a legend or direct annotations, and a caption area specifying sample size and confidence convention; a figure must not rely on its legend alone.
+6. Ablations and limitations are mandatory slides, not optional ones: ablations isolate each component’s contribution; limitations accurately state scope, failure modes, and ethics/de-identification status.
+7. Every number must include its unit, time range, dataset, and statistical convention (mean ± variance, sample size); express uncertainty using a measurable quantity or confidence interval, never vague language.
+8. Open with the problem statement and a one-sentence contribution; close with numbered conclusions (claims, not a summary) and an explicit list of open questions.
+9. Place citations in a compact footnote line at the bottom of the slide where they are used; unless the genre requires it, avoid a full-slide wall of dense references.
+
+Default baseline (distilled from the layout skill; PART B may override): organize each slide around one evidence unit (figure/table/equation); charts must include complete axis labels, units, and sample definitions; keep charts honest—do not truncate the y-axis to exaggerate differences, and distinguish correlation from causation explicitly in the copy; ensure projection readability on a 960×540 canvas, with body text at ≥14pt equivalent and chart labels at ≥10pt equivalent.
+
+**Title sequence test**: Reading only the slide titles in order should tell the complete story; if they do not form a coherent narrative, the structure is not working.
+
+Density anchor: The specific minimums for each style (number of evidence modules and charts per slide, annotation conventions) are defined by the 【Density Baseline】 reverse-engineered from that style’s source deck—this baseline does not impose a universal number.
+
+### Global Prohibitions (Default AI Patterns · Mandatory)
+
+1. **Do not use cards by default**: Unless the user explicitly requests them, never use rounded rectangles or rectangular cards to establish hierarchy or alignment—functional containers (status indicators, hypothesis boxes, quotation boxes) do not count as cards. Lines, white space, font size, and weight differences are preferable.
+2. **Do not use equal-split compositions**: Unless no other viable layout exists, do not default to three equal columns, four equal columns, or a two-by-two matrix. An equal split must be justified by the content structure (three scenarios naturally warrant three columns; frameworks explicitly named in the content are exempt).
+3. **Do not use unmotivated AI color schemes**: Institution-specific palettes stated in the content are exempt. Prohibited schemes are generic, familiar defaults with no rationale—blue-purple gradients, cyan-purple neon, rainbow glows, glass cards, and luminous borders.
+4. **Do not use elements that conflict with the overall style**: No out-of-style treatments, such as rounded icons or rounded rectangles in a sharp-edged visual system.
+5. **Keep chart backgrounds transparent**: Use transparency or the slide background color, except for reversed charts on dark backgrounds.
+6. **Do not leave planned regions empty**: Do not leave an entire block of the content area blank; every planned region must be filled with real content. Failure to fill it is the most serious form of roughness.
+
+## PART B — Signature System (Distinctive Features of This Template; Overrides PART A in Case of Conflict)
+
+One-sentence style signature: Editorial infographic courseware with a white background and deep ink text, using electric blue as the structural anchor, gray-blue/cyan for layered hierarchy, and an alternating rhythm of photography, evidence blocks, and geometric white space.
+
+【Color Palette】
+The background is predominantly `#FFFFFF`; body text and line art use `#202124`, reserved for high-contrast text; titles, top rules, numbering, key figures, and the primary panel use `#4285F4`, with only one large primary-blue block per slide; deep blue `#0059BA` is reserved for strong-result panels; cyan `#24C2E0` is used for secondary-option panels; light blue `#ADCCFA` is used for explanatory backgrounds; gray `#D8DBE0` is used for comparison, disabled, or decorative backgrounds, while `#9AA0A6` is used for headers and low-emphasis labels; positive or intervention markers use `#34A853`; do not introduce a separate red for negative or unaffected states—continue using `#202124` or `#9AA0A6`. Charts use a fixed mapping: primary blue = primary series, light gray = no occurrence, charcoal = control, cyan = secondary series. Use no more than 4 hues per slide; decorative elements use blue, black, and gray only; never flood the entire slide with an accent color.
+
+【Layout Skeleton】
+Source-deck facts: 1190.55×1683.78pt, A4 portrait, with an aspect ratio of approximately 0.7071 and left/right margins of approximately 80pt. Content slides have an approximately 2pt blue rule across the top, gray section navigation on the left, and a gray running title on the right; below the title, use either text on the left and a visual on the right or stacked horizontal bands, often with a geometric anchor in a bottom corner. Cover only: large title at top left; at the bottom, an arched/rounded photograph layered with a striped circle and black capsule. Section divider only: oversized blue number and title, with a large image at bottom right. Agenda slides may use vertical numbering. 16:9 adaptation: retain the top rule, navigation at both ends, left-text/right-visual structure, blue/gray bands, and bottom-corner geometry; convert vertical stacking into a horizontal 2-column layout or 3:2 partition, move the photograph to the right or into a lower horizontal band, and reduce circular decoration to a corner accent. Never present adaptation rules as source-deck facts.
+
+【Typography】
+Source-deck facts: sans serif `GoogleSans-Regular/Medium/Bold` (render with {latin: Liter, ea: MiSans}); body text is approximately 20pt=1x, argument titles 46pt≈2.3x, section/cover titles 80–87pt≈4–4.4x, large figures 60–70pt≈3–3.5x, running headers 18pt≈0.9x, and footnotes 14pt≈0.7x. Use Medium for titles, Regular for body copy, and Bold for metrics/conclusions; enlarge numbers while setting units and explanations one level smaller; keywords often use a blue underline. 16:9 adaptation: recommended, not measured—on a 960×540 canvas, use body text at 18–22pt equivalent, titles at 42–54pt, and section titles at 72–84pt, preserving the original scale and weight relationships.
+
+【Chart Language】
+Observed types: rows of person icons, layered tree/process diagrams, central-node loop diagrams, two-row treatment/control comparisons, small “pre–intervention–post” line charts, four-step numbered processes, and paired-metric case-study blocks. Dense conventional tables are not a stable signature; axis labels, units, and measurement conventions inherit PART A.
+
+⭐Signature techniques:
+1. Use gray/blue/black person icons with a matching legend to communicate group attribution.
+2. Use light blue → cyan → deep blue → gray nodes, thin black outlines, and orthogonal connectors to show a four-level hierarchy.
+3. Link line-art icons with a black loop line and place a primary-blue circle at the center to express system relationships.
+4. Inside a white rounded panel, use blue/gray line charts; use a green dashed line only to mark the intervention point.
+5. Arrange treatment/control as horizontal rows, with a device screenshot beside a thin divider to explain between-group differences.
+
+【Signature Components】
+Top rule + gray navigation: 1 instance on every content slide, fixed to the top edge. Geometric anchor cluster (striped circle, quarter circle, gray circle/square, black capsule/triangle): no more than 1 cluster per slide, attached only to an edge or corner. Blue quotation panel: no more than 1 per slide, with a large white quotation mark and a short quotation. Blue/gray case-study band: no more than 1 group per slide, layered with a device screenshot/photograph/large metric. Rounded action capsule: no more than 1 per slide and only for an explicit action. Line-art icons use black outlines plus only one blue/cyan accent. Photography is mandatory on narrative/case-study slides: use 1 hero image, cropped as a circle/arch/rounded shape/diagonal cut; never use a photo wall or replace it with an icon wall.
+
+【Prohibited】
+No gradients, neon, luminous borders, 3D charts, or diffuse shadows; no continuous sequence of rounded cards for ordinary paragraphs; do not tile striped circles/quarter circles as a background; no rainbow icons or multicolor gradient charts; no full-slide photographs or photo collages; oversized blue numbers are reserved for sections, steps, or key metrics; serif typography must not serve as the primary typeface.
+
+【Slide Types and Layouts】
+Section/Cover｜Opening transition｜Title 45% + photograph/geometry 55%｜2–4 blocks｜Large number, photograph, geometric cluster｜Do not include a complete evidence chain.
+Argument + Image/Quotation｜Claim with explanation/citation｜Left text 42–48% + right image/blue panel 52–58%｜4–7 blocks｜Photograph, quotation, key figure｜Do not include a multi-series statistical chart.
+Method/Process Diagram｜Steps, hierarchy, system relationships｜Diagram 45–55%, explanation 25–35%｜4–8 blocks｜Line art, colored nodes, numbering｜Do not include long-form narrative.
+Comparison/Two-Column Case Study｜Treatment/control or two cases side by side｜Blue/gray approximately 50% each｜6–10 blocks｜Device screenshot, paired metrics, action capsule｜Do not present a single continuous argument.
+Checklist/Do–Avoid｜Standards and review items｜Left/right approximately 45/45%, optional photograph along the bottom｜6–9 blocks｜Check/cross line art, gray background blocks｜Do not include a high-dimensional data table.
+
+【Density Baseline】
+Most content slides contain 1 main title and 4–8 independent blocks; two-column case studies contain 6–10 blocks; process slides contain 1 primary diagram + 3–5 explanatory elements. A slide usually contains 0–1 chart; complex slides may add at most 1–2 microcharts. Label only key figures and legends; do not crowd every bar or point with labels. Use 2–4 explanatory paragraphs, each approximately 40–90 words/characters equivalent; present metrics as a large number + unit/short conclusion. Concentrate white space in the header, between the title and primary visual, and around geometric elements in the bottom corners; the content area must still be fully occupied by substantive text, diagrams, or photography.

skills/open-kimi-ppt/reference/design_system/academic/deep-blue-atlas/design.md

@@ -0,0 +1,97 @@
+# Deep Blue Atlas · academic STYLE DESIGN SYSTEM
+
+## PART A — academic Baseline (Universal Floor)
+
+Academic presentations benchmarked against top-tier research talks and thesis defenses (NeurIPS/CHI/SIGGRAPH main-track caliber): clarity of argument outranks decoration; the whole deck must read as one reproducible scientific claim, not marketing collateral. Academic register throughout — the disciplined vocabulary of research questions, hypotheses, evidence, significance, contributions, and limitations; no marketing or business-speak mixed in.
+
+Content and layout:
+1. Fixed argument skeleton: research question → gap in related work → data and method → main results → ablation/analysis → limitations → future work; short decks may compress, but the order may not be reversed.
+2. Evidence first: results tables must include baseline comparison columns with the best values marked, relative gains beside absolute numbers; core contributions presented as formally typeset equations; method pages carry reproducibility information (data scale, annotation protocol, key hyperparameters).
+3. Medium information density: one claim per page, evidence organized around that claim; pages are designed, not filled — whitespace is a legitimate device for separating stages of the argument.
+
+Visual style:
+1. Aesthetics: plain, precise, scholarly; no slogans, decorative photography, or icon clip art.
+2. Body text and labels in a clear sans-serif; titles may use serif when the source style does; equations formally typeset (KaTeX-grade), with proper numbering and italicized variables.
+3. One two-color system: a primary structural color + one accent, grays carrying hierarchy; negative/ablation signals use a single consistent warning color.
+4. Diagrams in thin, precise vector lines, tables in minimal horizontal rules; charts carry axis labels, units, and sample basis; never truncate the y-axis to exaggerate differences; correlation and causation are explicitly distinguished in the copy; no 3D, shadows, or gradients as decoration.
+
+Structure and discipline:
+1. Every body-page title must be a claim (a finding, not a section name), with supporting evidence following on the page; every quantitative conclusion is backed on-page by a chart, table, or citation — no orphan numbers.
+2. Figures must be self-contained: axis labels with units, in-figure annotation or legends, captions giving sample size and confidence basis.
+3. Ablation and limitations pages are mandatory; numbers must carry units, time ranges, datasets, and statistical basis; uncertainty is expressed as confidence intervals.
+4. The opening is a problem statement plus a one-sentence contribution; the close is numbered conclusions plus a list of open questions; citations sit as compact footnote lines at the page bottom — no full-page reference walls.
+
+Global prohibitions (mandatory): no default cards (containers serving information functions do not count as cards); no evenly split compositions unless justified by content structure; no unmotivated AI color schemes (blue-purple gradients, neon, glass cards, glowing borders); no elements that clash with the overall style; chart backgrounds transparent or the page background color; care with large empty areas — and no padding just to fill space.
+
+Density anchor: per-page minimums for evidence modules, charts, and labeling follow the density-baseline section in PART B; this baseline imposes no uniform numbers.
+
+## PART B — Signature System
+
+One-line style signature: a deep-blue linear-reasoning system on white — oversized conclusory titles deliver the judgment first, orthogonal segments and nodes organize the evidence, cyan singles out the one key piece of evidence, and trace orange-red handles thresholds and exceptions.
+
+【Color Palette】
+All hex values below are approximations pixel-sampled from the reference images (marked "approx."):
+- Page background #FFFFFF (approx.): a uniform pure-white canvas across the deck, white covering over 90% of the area, no dark pages.
+- Primary structural deep blue #203D74 (approx.): page titles, body emphasis, table headers, key figures, chart main lines, nodes, arrows, and decision-tree edges — the system's first visual anchor.
+- Secondary deep blue #2C477A (approx.): section subheads and secondary structural lines, carrying the deep blue's anti-aliased tiers.
+- Primary accent cyan #49B7D0 (approx.): subtitles, two-digit numerals, selected/active states, point estimates, event nodes, current step; never used as large-area fill, and it doubles as the entire "positive/valid" semantics (this style uses no green).
+- Neutral gray #7F899E (approx.): explanatory text, axis labels, auxiliary notes, and legend-style annotations.
+- Light blue-gray #B5C0D0 (approx.): page numbers, secondary notes, de-emphasized information, and fine reference lines.
+- Divider gray #D9DEE7 (approx.): table hairlines, row separators, inter-column boundaries.
+- Grid gray #E7ECF1 (approx.): horizontal chart gridlines, faintest structural lines.
+- Negative red #E94B3B (approx.): inapplicable/excluded states, ex-ante threshold lines, risk reference lines — trace amounts only.
+- Reminder orange #FD9F66 (approx.): exceptions and flagged figures needing attention without a negative verdict — even more trace than red.
+Color discipline: no multi-series rainbow palette; within a page, no more than the three hue families deep blue / cyan / gray; every colored element must be able to explain its information role.
+
+【Layout Skeleton】
+Source facts: 16:9 canvas (sample images ~1467×825 px). Outer margins ~5–7% left/right, 6–8% top; page titles left-aligned, occupying roughly the top quarter; body content begins at ~28–32% of page height. The reading path is a Z: title → main figure/table → sidebar explanation → bottom conclusion line. Columns are always asymmetric — ratios such as ~58:42, 37:63, 31:38:31, 72:28 chosen per page type — with zones separated only by whitespace and fine gray rules, never panels. A light-gray "current page / total pages" number is fixed at lower right, omitted on the cover; no header navigation bar, no fixed sidebar, no logo corner mark. The bands between title zone and main figure, the inter-column transitions, and the page edges are fixed whitespace.
+
+【Typography】
+Source facts: a modern sans-serif character throughout — even stroke widths, flat terminals, open letterforms, close to a modern gothic/humanist sans (render with {latin: QuattrocentoSans, ea: MiSans}); Latin and numerals equally clean sans (the JPGs cannot confirm exact family names). Taking body text as 1x: small notes/page numbers 0.65–0.8x; section subheads 1.25–1.5x, semi-bold; page titles 2.1–2.5x, regular-to-medium weight, authority built through size rather than extreme boldness; cover main title 3–3.5x, distinctly bold; result big numbers 2–2.8x, semi-bold to bold. Weight hierarchy: page titles regular/medium, section heads semi-bold, key conclusions and figures bold, explanatory text regular, auxiliary information demoted to gray. No serif, script, italic, or monospace code faces anywhere in the set.
+
+【Chart Language】
+Only types that actually appear in the samples:
+- ⭐ Single-series line chart: light-gray horizontal grid, no legend, node values labeled directly, one in-chart sentence of annotation.
+- ⭐ Three-line academic wide table: only two deep-blue rules, below the header and at the foot — no vertical lines, no outer frame, no zebra striping; key-difference cells switch to cyan, variable differences legible without a legend.
+- ⭐ Orthogonal decision tree: diamond decision nodes + right-angle polylines + small terminal arrows, branch labels placed along the lines; leaf nodes = deep-blue bold name + status word (red = inapplicable, cyan = applicable) + gray explanation.
+- ⭐ Segmented timeline: thick deep-blue main axis + short ticks; excluded phases switch to gray dashed lines with axis-break jogs; cyan dots mark key instants, vertical leader lines connect annotations.
+- ⭐ Confidence-interval plot (simplified forest plot): deep-blue interval lines + end caps + cyan point estimates + gray reference dashes + a red ex-ante threshold line.
+- Mini interval scale: fine line with ticks at both ends, baseline value in gray at left, criterion value in deep-blue bold at right, flanking metric rows and logic rows.
+- Horizontal process chain: cyan-numbered steps joined by thin arrows, with one looping bracket line allowed to express retrospective iteration.
+Shared discipline: labels placed directly, no legend dependence; one main figure per page — never pile up multiple independent charts.
+
+【Signature Components】
+- Conclusory display title: one complete long-form assertion per body page, left-aligned, occupying the top quarter; on the cover, a main title plus core proposition performs the same function.
+- Two-digit cyan numerals: 01/02-style navigation numbers, small but conspicuous, used for list items and process steps — the deck's wayfinding.
+- Cyan evidence points: one to three cyan focuses per page (point estimate, key node, selected path, current step) — the reader's sole landing spot, never a fill.
+- Lines as grammar: dividers, axes, leader lines, and table rules carry all structural expression — the style's most stable visual fingerprint.
+- Panel-free zones: multi-column areas are zoned by whitespace and fine gray rules alone — no rounded cards, no shadows, no fill blocks.
+- Short-tick anchor lists: resolution/threshold lists use short cyan vertical ticks as per-item anchors instead of bullets.
+- Bottom conclusion line: one full-width deep-blue bold line at the foot of body pages, sealing the page's argument.
+- Light lower-right page number: light gray "current page / total pages", omitted on the cover.
+- Restrained geometric nodes: diamond = decision, dot = status/estimate, short tick = list anchor, arrow = direction; no general-purpose icon library.
+
+【Prohibited】
+- No dark full-bleed backgrounds; deep blue is for text and lines, never page fills.
+- No photos, people images, product shots, screenshots, illustrations, or large image backgrounds.
+- No gradients, glows, shadows, glassmorphism, transparency layering, or 3D effects.
+- No rounded cards, card walls, colored panels, or heavy information boxes.
+- No colored icons, emoji, stickers, badge-style labels, or decorative graphic assets; no logos, brand corner marks, permanent sidebars, or header navigation bars.
+- No vertical table lines, heavy outer frames, or zebra row fills; tables use horizontal rules only.
+- No pies, donuts, default office-suite legends, or multi-color categorical charts.
+- No green for positive states; positive/valid semantics are carried entirely by cyan.
+- No free-curve connectors, radial mind maps, or dense network node graphs — the "atlas feel" comes from orthogonal-line relationship expression, not webs.
+- No long centered copy, slogan-style one-liner pages, or decorative giant background numerals; no serif, script, italic, or display faces.
+
+【Slide Types and Layouts】
+Cover | module opening | left ~2/3 information zone + right 1/3 whitespace | eyebrow + oversized main title + cyan subtitle + one line of meta info + fine divider + bolded core proposition + supplementary note | ~7 text blocks, no charts, no page number.
+Problem evidence page | phenomenon data + misreading teardown | ~58:42 two columns | 1 main figure + 3 numbered items + bottom conclusion band | direct-labeled line chart, cyan numerals | not suited to parallel multi-topic content.
+Logic hypothesis page | falsifiable hypothesis statements | narrow left column with large relational words + wide middle column of explanation + mini interval plot at right | 3 logic rows + 2 dividers | not suited to long formula derivations.
+Comparison table page | groups/approaches compared side by side | full-width three-line table (~88% of page width) + basis note below + key-points passage + conclusion line | high density relieved by generous row height.
+Decision-tree page | option selection converging | ~37:63, conclusion first in the left column + orthogonal decision tree at right | 2–3 decision nodes + 4 leaf nodes | wide gaps between branches to avoid line crossings.
+Timeline page | measurement windows / phase schedule | full-width single axis (~86% of page width) + annotations above and below the axis + 2-line bottom note | medium-low density, breathing room around the axis.
+Results decision page | metric comparison + point estimates + action branches | ~31:38:31 three columns + bottom three-column action zone | the deck's density ceiling, order maintained by three-column alignment.
+Convergence page | scaling resolutions | ~72:28 main zone + right resolution rail | process chain + loop line + threshold checklist + limitations note | does not end on a thank-you page.
+
+【Density Baseline】
+One core visual object per page: 1 main figure/table/framework — never pile up multiple independent charts. Body pages carry ~7–15 text and graphic blocks; line charts directly label 3–5 nodes; comparison tables 2–4 data rows × 4–6 columns with generous row height; decision trees 2–3 decision nodes and at most 4 leaves; timelines carry at most 8 time-point labels; results pages allow 2 sets of mini metrics + 1 interval plot + 3 columns of action branches. Numbered lists run 3–5 items per page, each a one-line bolded claim + one gray explanatory sentence. Whitespace concentrates between title and main figure, in asymmetric-column transitions, and at the page edges; the body area holds no large empty block, nor may anything be stuffed in to fill. At most 3 cyan focuses per page; red and orange combined remain trace.

skills/open-kimi-ppt/reference/design_system/academic/paper-white-courseware/design.md

@@ -0,0 +1,66 @@
+# Paper White Courseware · Academic STYLE DESIGN SYSTEM
+
+## PART A — Category Baseline (General Minimum)
+
+Create rigorous, formally typeset academic presentations at the standard of top-tier research talks and thesis defenses (NeurIPS/CHI/SIGGRAPH main-stage quality), with clarity of argument taking precedence over decoration. The deck must read as a reproducible scientific claim, not as marketing collateral. Use academic language throughout: standard terminology for research questions, hypotheses, evidence, significance, contributions, and limitations, without marketing or business phrasing. Shift the visual emphasis by discipline: theoretical work should make equations and proof steps the dominant large-scale content; computer science and engineering should be structured around system diagrams, code, execution traces, and performance curves; experimental research should tightly organize equations, charts, and images around the same conclusion; humanities and social sciences should layer archival records, maps, manuscripts, and other textual evidence with narrative depth while preserving academic rigor.
+
+Content and layout:
+1. Use a fixed argument structure: research question → gap in related work → data and methods → main results → ablation/analysis → limitations → future work. A short deck may compress this sequence, but must not reorder it.
+2. Make every slide evidence-first: results tables must include baseline comparison columns; use comparison charts and architecture/process diagrams; present the core contribution as a formally typeset equation.
+3. Maintain medium information density: one claim per slide, with evidence organized around that claim. Slides should be composed, not filled—white space is a legitimate way to separate stages of the argument.
+
+Visual style:
+1. Aesthetic: restrained, precise, and scholarly. Do not use promotional copy, decorative photography, or clip-art icons.
+2. Typography: use clear sans serif typefaces (render with Liter / Chinese: MiSans) for body text and labels; titles may use serif type when the source style does. Equations must be formally typeset (KaTeX quality), with standard equation numbering and italic variables.
+3. Color: use a two-color system for each deck—one primary structural color and one accent color, with gray carrying hierarchy; use a white or near-white background; use only one consistent warning color for negative results or ablation signals.
+4. Graphics: draw diagrams with thin, precise vector lines; use minimal horizontal rules in tables; give charts axis labels, units, and sample definitions; do not use 3D effects, shadows, or gradients as decoration.
+
+Structure and discipline (derived from work we consider exemplary):
+1. Every content-slide title must state a claim—a finding, not a section label: “Lapses in attention can be predicted 10 seconds in advance,” not “Experimental results.” Support that claim with evidence on the same slide.
+2. Align claims with evidence: every quantitative conclusion on a slide must be supported there by a chart, table, or cited result. Never leave a number orphaned.
+3. The primary results table must include baseline comparison columns and mark the best value; show relative improvement alongside absolute values.
+4. Method slides must provide reproducibility details: dataset size, annotation protocol, inter-annotator agreement, and key hyperparameters. Show only the core idea as an equation; place derivations in appendix slides.
+5. Every figure must be self-contained: include axis labels with units, a legend or direct labels, and a caption area stating sample size and confidence convention. A figure must not rely on the legend alone.
+6. Ablations and limitations are mandatory, not optional: ablations isolate the contribution of each component; limitations candidly state scope, failure modes, and ethical/anonymization status.
+7. Every number must include its unit, time range, dataset, and statistical convention (mean ± variance, sample size). Express uncertainty with a measurable quantity or confidence interval, not vague language.
+8. Open with a problem statement and a one-sentence contribution; close with numbered conclusions—claims, not a summary—and an explicit list of open questions.
+9. Place citations as compact footnote lines at the bottom of slides where they are used; unless the genre requires it, do not create a dense full-slide wall of references.
+
+Default baseline (derived from the layout skill; PART B may override): organize each slide around one evidence unit (figure/table/equation); give charts complete axis labels, units, and sample definitions; make charts honest—do not truncate the y-axis to exaggerate differences, and distinguish correlation from causation explicitly in the copy; ensure projection readability, with body text on a 960×540 canvas at ≥14pt equivalent and chart labels at ≥10pt equivalent.
+
+**Title read-through test**: reading only the slide titles should tell the complete story; if the sequence does not flow, the structure is not working.
+
+Density anchor: the specific minimum for each style—the number of evidence modules and charts per slide, and annotation conventions—is defined by the 【Density Baseline】 reverse-engineered from that style’s source deck. This baseline does not impose a single universal number.
+
+### Global Prohibitions (Mandatory · Prevent Default AI Patterns)
+
+1. **No cards by default**: unless the user explicitly requests them, do not use rounded rectangles or rectangular cards to create hierarchy or alignment. Containers with a genuine information function—status indicators, hypothesis boxes, quotation frames—do not count as cards. Lines, white space, font size, and weight contrast are better tools.
+2. **No equal-split compositions**: unless no other layout is viable, do not default to three equal columns, four equal columns, or a two-by-two matrix. An equal split must be justified by the content structure—three scenarios naturally support three columns; frameworks explicitly named in the content are exempt.
+3. **No unjustified AI-default color schemes**: institution-specific palettes stated in the content are exempt. Prohibited defaults are generic, moderate, and unsupported choices such as blue-purple gradients, cyan-purple neon, rainbow flares, glass cards, and glowing borders.
+4. **No elements inconsistent with the overall style**: do not introduce forms outside the visual language, such as rounded icons or rounded rectangles in a sharp-edged style.
+5. **Keep chart backgrounds transparent**: use transparency or the slide background color, except for reversed charts on dark backgrounds.
+6. **No empty regions**: do not leave an entire planned region of the content area blank. Every allocated region must contain real content; an unfilled region is the most serious form of roughness.
+
+## PART B — Signature System (Distinctive Template Signature; This Section Overrides PART A in Case of Conflict)
+
+One-sentence style signature: warm courseware with a paper-editorial sensibility, grounded in creamy paper white; forest green establishes structure, coral orange provides small accents, and moderately high long-form text density coexists with generous breathing room.
+
+【Color Palette】Background #FDFAF5: fill the full slide while retaining the warmth of paper; primary text #56687A: use for all long-form text instead of pure black; title #44712E: use for slide titles, section labels, bold emphasis, and dark result bars; primary accent #F5987E: use only for short divider rules, chapter numbers, step numbers, and key figures; supporting green #D7EBCE: use only as fills for list items, table rows, or title panels; supporting pink #F9DED8: use only for chapter circles, case-study top bands, or alternating rows. Fix the chart series to #44712E, #F5987E, and their light variants #D7EBCE and #F9DED8. Do not create a separate red/green convention for positive and negative values; use #F5987E for both, with ↑/↓ and text indicating direction. Use no more than two hue families per slide (green + coral); never flood an entire slide with the accent color.
+
+【Layout Skeleton】Source-deck facts: 612×792 pt portrait Letter, with a 17:22 aspect ratio and standard margins of approximately 48 pt. Most content slides place a 25 pt claim title at the upper left, followed by a coral rule approximately 50×4 pt; the coral chapter number sits at the upper right, double angle brackets at the lower left, and “Short Document Name｜Page Number” at the lower right. Content typically uses two unequal columns, read left to right and then downward. Cover/chapter slides use only a giant pale-pink circle + a person or botanical illustration + a pale-green title panel at the bottom. For 16:9 adaptation: retain the left margin, short rule, upper-right chapter number, and footer; change the columns to approximately 45:55, use a horizontal photo band across the upper 35%–40%, enlarge the chapter slide’s large circle and shift it left, and convert the pale-green title panel into a wide lower-left band rather than stretching the portrait composition.
+
+【Typography】Source-deck font facts: the PDF embeds the slab-serif family Community Slab (Regular/Semibold/Italic) and the sans serif family Community (Light/Semibold); however, the reference output is entirely sans-serif with extra-bold black titles — the reference image wins: body, titles, numbers, and navigation all use the sans family (render with {latin: Liter, ea: MiSans}, titles and figures bold). Body text 11.5 pt = 1x, slide titles 25 pt ≈ 2.2x, information statements 15–17 pt ≈ 1.3–1.5x, large figures 40–48 pt ≈ 3.5–4.2x, chapter numbers approximately 105 pt ≈ 9x, and footers 8 pt ≈ 0.7x. For 16:9 adaptation: on a 960×540 canvas, use 16–18 pt body text, 34–40 pt slide titles, and 56–72 pt large figures; preserve the scale ratios and the all-sans weight hierarchy. Set English titles in sentence case, never all caps; bold only the conclusion within a sentence, and italicize quotations.
+
+【Chart Language】Elements used in the source: bar charts with few bars, single-diagonal relationship diagrams, funnels, circular-arrow diagrams, proportional icon arrays, comparison matrices/tables, and KPI result bars. ⭐Pair each chart with a one-sentence rule on the same slide: place the chart within a pale-green or pale-pink field, and label the conclusion directly beside the relevant line, bar, or region rather than creating a separate large legend. ⭐All quantified case-study outcomes belong in a dark-green result bar: white explanatory text + a large coral figure, with body content avoiding emphasis at the same level. ⭐Use proportional icon arrays only occasionally, and only when the sample is small enough for a proportion to be read directly from 10 units.
+
+【Signature Components】Coral short rule: use it once beneath the title, at a width approximately 2 times the title’s cap height; pale list panels: use only for parallel steps, comparison rows, or tool items, 3–6 per slide—this is a narrow exception to PART A’s prohibition on default cards; flat person/botanical illustrations: use them to turn abstract claims into situations, no more than one group per slide, positioned against the slide edge or in open space without covering body text; dark-green result bar: use only for “result + key change,” no more than one per slide. Photography is a hard requirement: include at least one real contextual photograph of a person in the deck, used only as a full-width top band or standalone rectangle on an introduction slide or at the opening of a case study. Do not replace it with an abstract gradient or an illustration-only treatment.
+
+【Prohibited】Do not use a photo as a text background or add a mask with overlaid text; do not generalize pale-green/pale-pink panels into a card grid with no information function; do not scatter colored icons—icons must sit inside white circular wells or be bound to list items; do not allow the native colors of embedded interfaces/screenshots to spill into the global palette; do not use coral orange for long passages of body text or as a full-slide background.
+
+【Slide Types and Layouts】Editorial argument slide｜one judgment + explanation + diagram｜title 15% / unequal two-column body 55% / marginal illustration or figure 30%｜3–5 blocks｜short rule, footer, marginal illustration｜not suitable for a multi-metric dashboard.
+Data-claim slide｜one large figure or one primary chart supports the conclusion｜title 15% / primary evidence 45% / explanation 30%｜2–4 blocks｜large figure, pale chart field｜not suitable for several small charts side by side.
+List/tool slide｜3–6 peer-level steps, benefits, or methods｜title 15% / vertical list or matrix 70%｜3–6 blocks｜pale list panels, icons in circular wells｜not suitable for a long derivation.
+Photo-led introduction slide｜human context + one-sentence lead-in｜photo 40%–48% / title and body 40%｜2–3 blocks｜full-width photo band, short rule｜not suitable for data-dense content.
+Case-evidence slide｜problem—method—result chain｜pale-pink top band or photo 20% / two-column body 50% / result bar 30%｜4–6 blocks｜device image or photo, dark-green result bar｜not suitable for a story without verifiable outcomes.
+
+【Density Baseline】Portrait argument slides in the source typically contain 3–5 evidence/explanation modules, list slides 3–6, and case-study slides 4–6; most slides contain 0–1 primary chart and never form a multi-chart dashboard. Label values bar by bar only in sparse bar charts with approximately 3 bars; label only the key regions and endpoints in relationship diagrams. For new 16:9 decks, use 2–4 modules per slide; keep each explanation block to roughly 40–90 characters’ worth of content, adjusted to an equivalent length in the working language. If a case study exceeds 4 blocks, split it into separate “Method” and “Results” slides. This overrides PART A’s prohibition on empty regions: continuous paper-white space is allowed beneath the title, around illustrations, and above the footer, but never reserve an empty container.

skills/open-kimi-ppt/reference/design_system/academic/pastel-derivation/design.md

@@ -0,0 +1,68 @@
+# Pastel Derivation · Academic STYLE DESIGN SYSTEM
+
+## PART A — Category Baseline (General Minimum Standard)
+
+A rigorous, formally typeset academic presentation benchmarked against top-tier research talks and thesis defenses (main-stage NeurIPS/CHI/SIGGRAPH quality), where clarity of argument takes precedence over decoration. The deck must read as a reproducible scientific claim, not as marketing collateral. Use academic language throughout, with disciplined wording for research questions, hypotheses, evidence, significance, contributions, and limitations; do not mix in marketing or business language. Shift the visual emphasis by field: theoretical derivations should devote most of the canvas to equations and proof steps; computer science and engineering should be structured around system diagrams, code, execution traces, and performance curves; experimental research should tightly organize equations, charts, and images around a single conclusion; humanities and social sciences should add narrative texture and textual evidence—such as archives, maps, and manuscripts—without compromising academic rigor.
+
+Content and layout:
+1. Use a fixed argument structure: research question → gap in related work → data and methods → main results → ablation/analysis → limitations → future work. A short deck may compress this structure, but must not change its order.
+2. Build evidence-first slides: include result tables with baseline comparison columns, comparison charts, and architecture/process diagrams; present the core contribution as a formally typeset equation.
+3. Maintain medium information density: one claim per slide, with evidence organized around that claim. Slides must be composed, not filled—white space is a valid way to separate stages of the argument.
+
+Visual style:
+1. Aesthetic: restrained, precise, and scholarly. Do not use promotional copy, decorative photography, or clip-art icons.
+2. Typography: use a clear sans serif (render with Liter / Chinese: MiSans) for body text and labels; headings may use a serif if the source style does. Equations must be professionally typeset (KaTeX quality), with standard equation numbering and italic variables.
+3. Color: use a two-color system for each deck—one primary structural color and one accent color—with gray establishing hierarchy. Use a white or near-white background. Use only one consistent warning color for negative or ablation signals.
+4. Graphics: use thin, precise vector lines for diagrams and minimal horizontal rules for tables. Charts must include axis labels, units, and sample definitions. Do not use 3D, shadows, or gradients as decoration.
+
+Structure and discipline (distilled from work recognized as benchmark quality):
+1. Every body-slide title must state a claim—a finding, not a section label: “Distraction Can Be Predicted 10 Seconds in Advance,” not “Experimental Results.” The evidence supporting that claim must follow on the same slide.
+2. Align claims with evidence: every quantitative statement on a slide must be supported on that slide by a chart, table, or cited result. No orphan numbers.
+3. The main results table must include baseline comparison columns and identify the best value; show relative improvement alongside absolute values.
+4. Method slides must include reproducibility details: dataset scale, annotation protocol, inter-annotator agreement, and key hyperparameters. Show only the central idea in equations; move derivations to appendix slides.
+5. Every figure must be self-contained: provide axis labels with units, a legend or direct labels, and a caption area specifying sample size and confidence basis. A figure must not rely on its legend alone.
+6. Ablations and limitations are required slides, not optional ones: ablations isolate the contribution of each component; limitations accurately state scope, failure modes, and ethical/anonymization status.
+7. Every number must include its unit, time span, dataset, and statistical basis (mean ± variance, sample size). Express uncertainty through measurable quantities or confidence intervals, not vague language.
+8. Open with the problem statement and a one-sentence contribution. Close with numbered conclusions—claims, not a summary—and an explicit list of open questions.
+9. Place citations in a compact footnote line at the bottom of the relevant slide. Do not create a dense, full-slide bibliography wall unless the format requires it.
+
+Default baseline (distilled from the layout skill; PART B may override): organize each slide around one evidence unit (figure/table/equation); give charts complete axis labels, units, and sample definitions; keep charts honest—do not truncate the vertical axis to exaggerate differences—and distinguish correlation from causation explicitly in the copy; ensure projection readability, with body text at ≥14pt equivalent and chart labels at ≥10pt equivalent on a 960×540 canvas.
+
+**Title read-through test**: reading only the slide titles in sequence should tell the complete story. If the sequence does not flow, the structure is not working.
+
+Density anchor: the specific minimum for each style—evidence modules per slide, chart count, and annotation conventions—is defined by the 【Density Baseline】 reverse-engineered from that style's source deck. This baseline does not impose one universal number.
+
+### Global Prohibitions (Mandatory · Prevent Default AI Patterns)
+
+1. **Do not default to cards**: unless the user explicitly requests them, do not use rounded rectangles or rectangular cards to create hierarchy or alignment. Containers with a genuine information function—such as status indicators, hypothesis boxes, or quotation boxes—do not count as cards. Lines, white space, type size, and weight differences are preferable.
+2. **Do not default to equal-split compositions**: unless no other layout is viable, do not use three equal columns, four equal columns, or a two-by-two matrix by default. Equal division must be justified by the content structure (three scenarios naturally warrant three columns; a framework explicitly named in the content is exempt).
+3. **Do not use ungrounded AI color schemes**: an institutional palette specified in the content is exempt. Prohibited choices are generic, commonplace, and unjustified defaults—blue-purple gradients, cyan-purple neon, rainbow light blobs, glass cards, and glowing borders.
+4. **Do not use any element inconsistent with the overall style**: styles from outside the system are prohibited, such as rounded icons or rounded rectangles in a sharp-edged visual language.
+5. **Keep chart backgrounds transparent**: use transparency or the slide background color, except for reversed charts on dark backgrounds.
+6. **Do not leave regions blank**: do not leave an entire planned region empty within the body content area. Every planned region must contain real content; failing to fill it is the most serious sign of an unfinished layout.
+
+## PART B — Signature System (Distinctive Template Signature; PART B Takes Precedence in Conflicts with PART A)
+
+One-line style signature: approachable, rational, whiteboard-like derivations on white, with pale pink–mist blue–mint chapter colors; electric blue and magenta carry fixed variable semantics; low-to-medium density with progressive reveal.
+
+【Color Palette】Body background: #FFFFFF. Use low-contrast gradients only on cover and chapter slides; observed corner colors are approximately #FEF2E8 at top left, #DBEDFE at top right, and #DEF8E9 at bottom right. Use #172C34 for all headings and readable text. Use the primary accent #0064E0 only on one key action phrase within a title. Use annotation blue #0085FD for arrows, links, newly introduced variables, and in-figure labels. Use supporting magenta #C32E8F only for the inverse operation, source term, or limiting phrase paired with blue. Use light gray #F1F4F7 only as a background frame for theorems, conclusions, or conditions, covering no more than 25% of a slide. Use almond #FCEEDC only as a “sandbox” for geometry, distributions, and trajectories. Chart series use #1F77B4, #FF7F0E, #2CA02C, and #D62728; use #0085FD for positive signals and #C32E8F for negative signals/caveats, with each applied to no more than a short phrase. Except in multi-series charts, use no more than two chromatic hues per slide. Pastels do not enter body-slide backgrounds, and accent colors never flood the slide.
+
+【Layout Skeleton】Source canvas: 1920×1080, 16:9. Anchor body-slide titles at the top left, starting at approximately x=2.7%, y=2.8%, and occupying 9%–12% of the height. Begin content at y≈15%; reserve the bottom 4% for bright-blue citations, with no full-width footer. The default reading path is “left-side definition/starting point → central arrow or main equation → right-side geometry/result.” When a horizontal flow is unsuitable, use a centered equation with a conclusion strip at the bottom. Primary evidence occupies 55%–70%; supporting explanation occupies 25%–40%. Do not use equal-weight cards. The cover alone uses a “large center-left title + right-side call to action + contributor information at the bottom” layout. Chapter slides alone use a slightly left-of-center “number + chapter title” layout, with a large QR code on the right only when a genuine external entry point exists.
+
+【Typography】PDF embedding has been verified: use OptimisticDisplay-Medium for headings, with a single key phrase optionally switched to OptimisticDisplay-Bold. Use OptimisticText-Regular/Bold for body text and labels. Use STIXGeneral-Regular/Italic for equations, with italic variables and upright operators. Taking the source deck's body text size of 40 as 1x, the heading size of 74 is approximately 1.85x. There is no consistent large-numeral tier. Main equations are usually 1.4–1.75x, with occasional use of 2.8x on minimal mechanism slides; figure captions are 0.6–0.75x and citations 0.5x. Use sentence case for headings, never all caps. Serif type is reserved for mathematics. Render with {latin: Liter, ea: MiSans} (headings Medium/Bold); use STIX Two Text for equations (via customFonts).
+
+【Chart Language】The source uses single- and dual-line charts, paired bar comparisons, minimal comparison tables, sample matrices, scatter plots, vector fields, trajectory schematics, and process-node diagrams. Tables inherit PART A, with no consistent additional treatment. ⭐Almond sandbox: use contours, point clusters, fine trajectories, and a small number of samples to spatialize abstract objects. ⭐Color-linked equation and diagram: key equation terms share their color with the corresponding arrows, points, and labels in the figure. ⭐Fixed-base progression: retain the same base figure and viewpoint across 2–5 adjacent slides, adding only one trajectory, condition, or equation term per slide; only the new element receives color. Standard performance charts do not label every data point; retain axes, legend, and source.
+
+【Signature Components】①Light-gray theorem/conclusion strip: spans the lower half of the slide or sits next to the relevant equation; use no more than 2 per slide, and only for a theorem, condition, conclusion, or caveat. ②Almond canvas: uses a 1–2 px dark outline; use 1–3 per slide, and only for geometry, distributions, trajectories, and corresponding samples. ③Blue downward arrow + magenta upward arrow: expresses only inverse operations or two directions of operation; use no more than 1 pair per slide. ④Within-chapter route strip: place 3 thumbnails side by side; add a #F1F4F7 backplate to the current stage and reduce the others to approximately 40% opacity. Use only on subsection transition slides. ⑤Progressive construction sequence: established graphics remain fixed in place; each new slide reveals only one new piece of evidence or transformation. Do not use this pattern for a single-slide summary.
+
+【Prohibited】Do not use pastel gradients, almond, or light gray as a full-slide background for ordinary body slides. Do not use serif type for explanatory prose. Do not turn #F1F4F7 rounded frames into generic cards; every frame must serve a theorem, condition, or conclusion. Do not use blue or magenta as decoration without semantic meaning, or swap their roles within the same progressive sequence. Do not expand a complete multi-step derivation on a single slide; split it across progressive slides while keeping the base figure fixed. Do not use reversed text on a dark background for body slides.
+
+【Slide Types and Layouts】
+Equation–Conclusion Slide | Definitions, losses, theorems, equivalences | Title 12% + central equation 60% + bottom conclusion 20% | 1–2 blocks | Light-gray conclusion strip, color-linked variables | Unsuitable for photo matrices.  
+Schematic Derivation Slide | Mechanisms or paths that map algebraic concepts | Almond figure 55%–65% + equation/explanation 35%–45% | 1–3 blocks | Almond canvas, paired arrows, progressive sequence | Unsuitable for dense literature.  
+Comparative Evidence Slide | Comparison of methods, conditions, or results | Two columns at 48/48 or three evidence units in a row | 2–4 blocks | Blue phrase in title, light-gray conclusion strip | Unsuitable for long equation chains.  
+Within-Chapter Navigation Slide | Subtopic progress | Three thumbnails across the horizontal middle band | 3 blocks, 1 current state | Within-chapter route strip | Does not carry new evidence.  
+References/Limitations Slide | Literature, open questions, or caveats | 1–3 light-gray functional boxes stacked vertically | 1–3 blocks | Bright-blue citations, magenta limiting phrases | Unsuitable for main results.  
+Occasional · Proposition Pause Slide | Problem reframing or within-chapter turn | Large proposition on the left 40% + geometric metaphor on the right 55% | 2 blocks | Almond geometric figure | Unsuitable for dense data.
+
+【Density Baseline】The source deck has low-to-medium density because it relies on progressive construction; no deck-wide hard minimum applies. An Equation–Conclusion Slide usually contains 1 main equation + 0–1 conclusion strip and 0–2 explanatory blocks, with a 15–50 words/characters equivalent per block. A Schematic Derivation Slide contains 1–3 figures and adds only 1 key node per slide; annotations follow variables and trajectories rather than labeling every data point. A Comparative Evidence Slide contains 2–4 modules, 0–2 quantitative charts, and optionally 1 conclusion; it uses 1–2 explanatory blocks, with a 25–60 words/characters equivalent per block. A References/Limitations Slide contains 2–3 boxes, with 3–6 lines per box. Preserve white space between the title and primary evidence, between two layers of equations, and around large figures. When content exceeds capacity, add slides for progressive reveal; do not shrink type to fill white space.

skills/open-kimi-ppt/reference/design_system/academic/teal-green-academic-defense/design.md

@@ -0,0 +1,90 @@
+# Teal-Green Academic Defense · Academic STYLE DESIGN SYSTEM
+
+## PART A — Category Baseline (Universal Minimum)
+
+Create a rigorous, formally typeset academic presentation benchmarked against top-tier research talks and thesis defenses (at the level of the main programs of NeurIPS/CHI/SIGGRAPH), with clarity of argument taking precedence over decoration. The deck must read as a reproducible scientific argument, not as marketing collateral. Use academic language throughout: standard terminology for research questions, hypotheses, evidence, significance, contributions, and limitations, with no marketing language or corporate jargon. Shift the visual emphasis by discipline: theoretical work should devote substantial space to equations and proof steps; computer science and engineering should be structured around system diagrams, code, execution traces, and performance curves; experimental research should organize equations, charts, and images tightly around a single conclusion; humanities and social sciences should layer archival material, maps, manuscripts, and other textual evidence with narrative depth on top of academic rigor.
+
+Content and layout:
+1. Use a fixed argumentative structure: research question → gap in related work → data and methods → main results → ablation/analysis → limitations → future work. A short deck may compress this structure, but must not change the sequence.
+2. Build evidence-first slides: result tables must include baseline comparison columns, comparative charts, and architecture/process diagrams; present core contributions as formally typeset equations.
+3. Maintain medium information density: one claim per slide, with evidence organized around that claim. Slides should be composed, not filled—whitespace is a valid way to separate stages of the argument.
+
+Visual style:
+1. Aesthetic: restrained, precise, and scholarly. Do not use slogans, decorative photography, or clip-art icons.
+2. Typography: use a clear sans serif (render with Liter / Chinese: MiSans) for body text and labels; headings may use a serif when the source style does; equations must be formally typeset (KaTeX quality), with standard equation numbering and italic variables.
+3. Color: use a two-color system for each deck—one primary structural color and one accent color, with gray establishing hierarchy; use white or near-white backgrounds; use only one consistent warning color for negative findings or ablation signals.
+4. Graphics: use thin, precise vector lines for diagrams; minimal horizontal rules for tables; axes, units, and sample definitions for charts; do not use 3D effects, shadows, or gradients as decoration.
+
+Structure and discipline (distilled from work established as the benchmark):
+1. Every content-slide title must state a claim (a finding, not a section label—“Attention Lapses Can Be Predicted 10 Seconds in Advance,” not “Experimental Results”), followed by evidence that supports the claim.
+2. Align claims with evidence: every quantitative statement on a slide must be supported by a chart, table, or cited result on the same slide; orphan numbers are not allowed.
+3. The main results table must include baseline comparison columns and identify the best value; show relative improvement alongside absolute values.
+4. Method slides must include reproducibility details: dataset size, annotation protocol, inter-annotator agreement, and key hyperparameters. Show only the core idea as an equation; place derivations in appendix slides.
+5. Every chart must be self-contained: include axis labels with units, a legend or direct labels, and sample size/confidence definitions in the caption area; a chart must not rely on its legend alone.
+6. Ablation and limitations are mandatory slides, not optional ones: ablation isolates the contribution of each component; limitations candidly state scope of applicability, failure modes, and ethics/de-identification status.
+7. Every number must include a unit, time range, dataset, and statistical definition (mean ± variance, sample size); express uncertainty through measurable quantities or confidence intervals, not vague wording.
+8. Open with the problem statement and a one-sentence contribution; close with numbered conclusions (claims, not a summary) and an explicit list of open questions.
+9. Place citations in a compact footnote line at the bottom of the slide where they are referenced; unless the format requires it, do not create a full slide densely packed with references.
+
+Default baseline (distilled from the layout skill; PART B may override): organize each slide around one evidence unit (chart/table/equation); give charts complete axis labels, units, and sample definitions; represent data honestly, without truncating the y-axis to exaggerate differences; distinguish correlation from causation explicitly in the copy; ensure projection readability, with body text at ≥14 pt equivalent and chart labels at ≥10 pt equivalent on a 960×540 canvas.
+
+**Title sequence test**: Read only the slide titles from beginning to end; they should tell the complete story. If the sequence is incoherent, the structure has not been established.
+
+Density anchor: The specific minimum for each style (number of evidence modules and charts per slide, annotation conventions) is defined by the 【Density Baseline】 reverse-engineered from that style’s source deck—this baseline does not impose a universal number.
+
+### Global Prohibitions (Default AI Patterns · Mandatory)
+
+1. **Do not default to cards**: Unless explicitly requested by the user, do not use rounded rectangles or rectangular cards to establish hierarchy or alignment—containers with a clear information function (status indicators, hypothesis boxes, quotation frames) are not considered cards. Lines, whitespace, font-size differences, and weight differences are preferred.
+2. **Do not use equal-part layouts by default**: Unless no other viable layout exists, do not default to three equal columns, four equal columns, or a two-by-two matrix—equal division must be justified by the content structure (three scenarios naturally support three columns; frameworks explicitly named in the content are exempt).
+3. **Do not use ungrounded AI color schemes**: Institution-specific palettes defined in the content are exempt; what is prohibited is a bland, common, unjustified default palette—blue-purple gradients, cyan-purple neon, rainbow flares, glass cards, and glowing borders.
+4. **Do not use elements that conflict with the overall style**: No out-of-style elements, such as rounded icons or rounded rectangles in an angular visual system.
+5. **Keep chart backgrounds transparent**: Use transparency or the slide background color, except for reversed charts on dark backgrounds.
+6. **Do not leave planned regions empty**: Do not leave an entire planned region of the content area blank; every planned region must contain real content. Failure to fill one is the most serious form of roughness.
+
+## PART B — Signature System (Template-Specific Signature; This Section Prevails over PART A in Case of Conflict)
+
+One-sentence style signature: Academic diagrams with black text on white, using teal green as the structural color, sky blue as the framework color, and yellow-orange as a single-point marker, balanced by medium-density whitespace.
+
+【Color Palette】
+
+Background #FFFFFF; use near-white #F7F7F4 only inside chapter-title frames. Primary text and headings #000000. Primary accent #009682 for square bullets, key connectors, checkmarks, and small labels, covering no more than 10% of any slide. Supporting colors: #68BCE0 only for goal banners/column headers, #939FB5 only for sidebars/inactive states, #868686 only for neutral lines, #70D6A6 for evidence labels, and #F4FAFD for goal-cell backgrounds. Focal marker #F0A81D may circle only one dimension or parameter. Chart series use #1F77B4, #FF7F0E, #2CA02C, #D62728, and #9467BD. Positive #009682; error/risk approximately #B51E00. Use 3–4 structural hues per slide; never flood a slide with an accent color.
+
+【Layout Skeleton】
+
+The source canvas is 483.874×272.126 pt, with a 1.778 (16:9) aspect ratio. On standard slides, place the title at the upper left, approximately 4% from the left and 12% from the top, in bold black; begin the body at approximately 26% of the slide height, using “left list + one chart/table/process diagram on the right,” with 38–45% for the left and 45–52% for the right. On process slides, center the main diagram horizontally; on chart slides, let one evidence object occupy the primary area. Standard slides include bottom navigation: section name + dot progress, with the current section bold black; on the next line, use a thin gray rule, page number/date/short title. Chapter slides contain only a centered rectangle with a thin teal-green outline and near-white interior, with no body copy. Title → key points → evidence → annotation.
+
+【Typography】
+
+The source deck’s primary typeface is the sans serif NimbusSanL-Regu/Bold (render with Liter / Chinese: MiSans); use mathematical/monospaced faces for equations and code (render with JetBrains Mono, via customFonts). Measured body text is 8.52 pt=1x, slide titles 13.63 pt≈1.60x, chapter titles 23.55 pt≈2.76x, footnotes/footers 5.68 pt≈0.67x; subsection bold text is 9.46 pt, banner text 11.36 pt. At 960×540, the proportional equivalents are approximately 17/27/47/11 pt. Slide titles commonly follow “Primary Class - Subclass”; use sentence case for body copy, uppercase for proper names/acronyms, and keep numbers, Greek letters, superscripts, subscripts, and equations aligned on the same line. Use bold or a word-level background highlight for emphasis.
+
+【Chart Language】
+
+The actual chart and diagram types include horizontal process diagrams, document/database pipelines, line charts, same-scale small multiples, horizontal bars, scatter plots + error bars, heatmaps, ablation tables, and comparison tables. Signature feature 1: Split the main visual into 2–6 same-scale panels with shared coordinates. Signature feature 2: Use top and bottom horizontal rules in tables, bold the best value, and place footnotes close to the bottom. Signature feature 3: Use thin black outlines, gray numbered bands, and one-way arrows in process diagrams. Signature feature 4: Direct-label series/key points within trend charts; retain axes, units, and sample definitions.
+
+【Signature Components】
+
+1. Document-to-output mapping diagram: folded-corner paper with black outlines, gray text lines, and sparse color highlights, with 1–3 light-gray curved lines connecting output icons; at most one set per slide. One purple-gray outlined callout bubble may explain a single piece of evidence.
+2. Goal banner/matrix: a sky-blue rounded banner with a line-art diamond at the left; below it, add a gray baseline/gap band and four columns of pale-blue cells, with a blue-gray vertical sidebar and checkmarks/plus signs where needed. Use only for goals, stage checks, and completion status.
+3. Evidence labels: thin-outlined teal-green rounded labels beneath the chart, with a document icon on the left and a short bold name/number on the right; at most two per slide.
+4. Navigation footer: section names, dot progress, page number, and date; show the current section in black and the rest in light gray.
+5. Bullets: use solid teal-green squares in different sizes for the first and second levels, with bold group headings; do not mix them with circles or checkmarks.
+
+【Prohibited】
+
+1. Do not use large teal-green/sky-blue backgrounds, gradients, shadows, or glass effects; sky blue is reserved for goal banners/column headers.
+2. Do not use rounded cards without an information purpose, four-way card walls, or heavy borders; reserve rounded corners for the three functional component types.
+3. Do not use solid-color chapter slides; chapter transitions must use a thin teal-green outlined rectangle, while content-slide titles remain at the upper left.
+4. Do not omit the navigation footer from standard content slides; chapter covers, matrix openers, and acknowledgments are exceptions.
+5. Do not place multiple high-saturation accent colors side by side; yellow marks only one focal point, red marks only errors/risks, and line art must not become skeuomorphic illustration.
+
+【Slide Types and Layouts】
+
+Narrative summary slide | problem/method/result highlights | 42% list on the left, 48% chart/screenshot on the right | 2–4 blocks | square bullets, document diagram | unsuitable for long tables.
+Process method slide | reproducible processing pipeline | horizontal 3–6 nodes across the center, with a numbered band along the top edge | 1 main diagram + 3–6 nodes | gray numbered band, line art | unsuitable for stacking conclusions.
+Results chart slide | trends, comparisons, ablations | one chart/table occupies 45–60%, with 2–3 interpretation points on the left | 1 primary evidence object | fixed series colors, best value in bold | unsuitable for a second large chart.
+Goal matrix slide | multidimensional goals and stage acceptance | banner at approximately 10% across the top, four-column grid + left vertical sidebar | 4 columns, 4–5 status rows | sky-blue banner, blue-gray sidebar | unsuitable for free-form narrative.
+Chapter transition slide | start of a new section | only a centered outlined title frame with generous whitespace on all sides | 1 title block | teal-green outlined rectangle | unsuitable for body evidence.
+
+【Density Baseline】
+
+A standard “list + visual” slide contains 2–4 blocks: 2–3 groups of key points on the left and 1 chart/table on the right; each group contains 2–4 bullets, each approximately 8–22 words/characters equivalent. A process slide contains 1 main flow, 3–6 nodes, and 0–1 callout. A chart slide typically contains 1 chart/table; same-scale small multiples contain 2–6 panels, with labels only on key bars/points. A matrix slide uses a fixed four-column grid, with 1–2 short statements or 1 status symbol per cell and no more than 5 rows. Explanatory blocks span 2–4 lines, each approximately 12–28 words/characters equivalent; place 1–2 lines of footnotes close to the chart or slide bottom. Preserve whitespace between the title and body, around the main visual on the right, and throughout chapter transition slides.

skills/open-kimi-ppt/reference/design_system/academic/wine-red-data/design.md

@@ -0,0 +1,62 @@
+# Wine Red Data · Academic STYLE DESIGN SYSTEM
+
+## PART A — Category Baseline (General Minimum Standard)
+
+A rigorous, formally typeset academic presentation benchmarked against top-tier research talks and thesis defenses (at the standard of NeurIPS/CHI/SIGGRAPH main conferences), where clarity of argument takes precedence over decoration. This deck must read as a reproducible scientific claim, not as marketing collateral. Use academic language throughout: standard terminology for research questions, hypotheses, evidence, significance, contributions, and limitations, with no marketing language or business jargon. Shift the visual emphasis by discipline: theoretical work should devote substantial space to equations and proof steps; computer science and engineering should be structured around system diagrams, code, execution traces, and performance curves; experimental research should tightly organize equations, charts, and images around a single conclusion; and humanities and social sciences should layer archival materials, maps, manuscripts, and other textual evidence with a sense of narrative, without compromising academic rigor.
+
+Content and layout:
+1. Use a fixed argumentative structure: Research Question → Gap in Related Work → Data and Methods → Main Results → Ablation/Analysis → Limitations → Future Work. A short deck may compress this structure, but must not change the sequence.
+2. Evidence-first slides: results tables must include baseline comparison columns, comparison charts, and architecture/process diagrams; core contributions must be expressed through formally typeset equations.
+3. Moderate information density: one argument per slide, with evidence organized around that argument. Slides should be composed, not filled—whitespace is a legitimate way to separate stages of the argument.
+
+Visual style:
+1. Aesthetic: plain, precise, and scholarly. Do not use promotional copy, decorative photography, or clip-art icons.
+2. Typography: use a clear sans serif for body text and labels (render with Liter / Chinese: MiSans); titles may use a serif when the source style does. Equations must be formally typeset (KaTeX quality), with correct equation numbering and italic treatment for variables.
+3. Color: use a two-color system for each deck—one primary structural color plus one accent, with gray carrying hierarchy; use a white or near-white background; use only one consistent warning color for negative results or ablation signals.
+4. Graphics: use thin, precise vector lines for diagrams; minimal horizontal rules for tables; and axis labels, units, and sample definitions for charts. Do not use 3D, shadows, or gradients as decoration.
+
+Structure and discipline (derived from work we recognize as benchmark quality):
+1. Every content-slide title must state an argument—a finding, not a section label: “Attention Lapses Can Be Predicted 10 Seconds in Advance,” not “Experimental Results.” The evidence supporting that argument follows on the same slide.
+2. Claim–evidence alignment: every quantitative conclusion on a slide must be supported on that same slide by a chart, table, or cited result. No orphan numbers.
+3. The main results table must include baseline comparison columns and mark the best value; show relative improvement alongside absolute values.
+4. Method slides must include the information needed for reproducibility: data scale, annotation protocol, inter-annotator agreement, and key hyperparameters. Show only the core idea as an equation; move derivations to appendix slides.
+5. Every figure must be self-contained: include axis labels with units, a legend or direct labels, and a caption area stating sample size and confidence basis. Figures must not rely on legends alone.
+6. Ablations and limitations are mandatory slides, not optional ones: ablations isolate the contribution of each component; limitations candidly state scope of applicability, failure modes, and ethics/de-identification status.
+7. Every number must include its unit, time range, dataset, and statistical basis (mean ± variance, sample size); express uncertainty through observable quantities or confidence intervals, never vague wording.
+8. Open with the problem statement plus a one-sentence contribution; close with numbered conclusions—claims, not a summary—and an explicit list of open questions.
+9. Citations appear as compact footnote lines at the bottom of the slides where they are referenced; do not create a full-slide wall of dense references unless the genre requires it.
+
+Default baseline (derived from the layout skill; PART B may override): Organize each slide around one evidence unit (figure/table/equation); give charts complete axis labels, units, and sample definitions; keep charts honest by avoiding truncated vertical axes that exaggerate differences, and distinguish correlation from causation explicitly in the copy; ensure projection readability, with body text at ≥14pt equivalent and chart labels at ≥10pt equivalent on a 960×540 canvas.
+
+**Title read-through test**: Reading only the slide titles should tell the complete story. If the sequence does not flow, the structure has not been established.
+
+Density anchor: The specific minimum standard for each style—the number of evidence modules and charts per slide, plus annotation conventions—is defined by that style’s 【Density Baseline】 reverse-engineered from the source deck. This baseline does not impose a uniform number.
+
+### Global Prohibitions (Default AI Patterns · Mandatory)
+
+1. **No default card layouts**: Unless the user explicitly requests them, never use rounded rectangles or rectangular cards to create hierarchy or alignment. Containers with a genuine information function—status indicators, assumption boxes, and quotation frames—do not count as cards. Rules, whitespace, type size, and weight contrast are preferable.
+2. **No equal-split compositions**: Unless no other layout is viable, do not default to equal thirds, equal quarters, or two-by-two matrices. Equal splits must be justified by the content structure—three scenarios naturally warrant three columns; frameworks explicitly named in the content are exempt.
+3. **No arbitrary AI-default color schemes**: Institutionally specified palettes in the body are exempt. What is prohibited is a generic, commonplace, unjustified default palette—blue-purple gradients, cyan-purple neon, rainbow glows, glass cards, and luminous borders.
+4. **No elements inconsistent with the overall style**: Do not introduce out-of-style treatments—for example, rounded icons or rounded rectangles in a sharp-edged visual system.
+5. **Keep chart backgrounds transparent**: Use transparency or the slide background, except for reversed-out charts on dark backgrounds.
+6. **No empty regions**: Do not leave large blank areas in the body-content region. Every planned area must be filled with real content; an unfilled region is the most serious form of roughness.
+
+## PART B — Signature System (Template-Specific Signature; This Section Prevails in Case of Conflict with PART A)
+
+One-sentence style signature: predominantly white-background academic lecture notes, with black type accented by wine red and purple-orange; full-slide solid-color chapter transitions; and information consistently resolved into “argument—evidence—source.”
+
+【Color Palette】The source deck uses “white-background content + sectional reskinning.” Background #FFFFFF (approximately 108/150 slides); chapters may use full-slide #820000 wine red (approximately 7 slides); covers and contents pages often use #F3F3F3 (approximately 13 slides); a standalone subsection later in the deck uses a #FFEFD9 cream background with #004032 dark-green text (approximately 13 slides), never mixed with wine red on the same slide. Primary text #000000; title purple #4F0E86 is reserved for openings/navigation; primary accent orange #E69138 is used for keywords, numbers, and arrows and occupies no more than 10–15%; fluorescent yellow #F7D05B (approximate) is used only for short horizontal brush strokes. Common chart-series colors are #0076BA, #00A2FF, #56C1FF, #D2C7EB, and #A9AAAC, with gray as the baseline; positive #00B050 and negative #FF0000 convey status only. Keep table backgrounds white or light gray; do not invent additional colored fills. Use 2–3 hues per slide, and never let the accent color fill the entire slide.
+
+【Layout Skeleton】Source-deck specifications: 720×405 pt, approximately 1100×619 px, 16:9. Place content-slide titles at top left with a left margin of approximately 4–6%. The body typically uses a 40/60 or 50/50 two-column split, with the main chart, screenshot, definition box, or table in the right column. Separate modules with whitespace and fine rules; a single gray source line often runs along the bottom, while slide numbers are not fixed throughout the deck. Special slides include centered chapter titles on light gray/light purple, full-slide declarations in wine red/black/dark green, and numbered contents pages whose current item is marked with a yellow brush stroke or colored capsule; these treatments are for transitions only. The reading path is title conclusion → chart/process → explanation → source. Reskinning applies continuously across an entire section and must not be mixed within a single slide.
+
+【Typography】PDF measurements show section-specific combinations: standard slides primarily use ArialMT/Arial-BoldMT, with OpenSans/Roboto appearing at the beginning; use Consolas for code; and SpaceGrotesk + Jomolhari in the later cream-colored subsection. When only slide images are available, do not guess font families; preserve the sans-serif/serif relationship. Body = 1x, titles 1.6–2x, large chapter text 2.3–3.4x, large numbers 1.4–2x, and footnotes 0.5–0.7x. Set titles in bold; keywords may use bold or italic orange/red. Use sentence case for body copy and all caps for short labels when appropriate; keep numbers and English abbreviations aligned to the baseline, and preserve formal equation typesetting. Render with {latin: Liter, ea: MiSans}; use JetBrains Mono for code (via customFonts).
+
+【Chart Language】Observed in the source: bar/line/scatter charts and dot distributions, radar charts, comparison tables, equation-definition tables, closed-loop processes, and screenshot comparisons. ⭐Signatures: ① The title states the conclusion first, and the chart only proves it; use one orange or wine-red series as the focal point and render the rest in gray/low-saturation blue (20+ slides). ② Show methods as horizontal steps or closed-loop arrows, with lightly filled nodes and thin black lines (15+ slides). ③ Place definitions/quotations in light-gray or pale-yellow rounded boxes beside the main figure (10+ slides). ④ Use human figures or dot arrays converging on a conclusion box to represent population distributions (approximately 8 slides). ⑤ High-density tables use only fine horizontal rules, equation columns, and a source line at the bottom (approximately 6 slides).
+
+【Signature Components】Apply the yellow brush stroke only to the current contents item or a key short phrase; keep it short and horizontal (approximately 3–4 instances). A full-slide wine-red declaration uses large white type for transitions and carries no data (approximately 7 slides). Rounded quotation/definition boxes hold direct quotations, measurement definitions, or status, with no more than 2–3 per slide. Red/orange outlines, strike-through lines, and arrows attach only to the marked object. Human/robot and similar icons serve only as process nodes or population glyphs. Photographs/screenshots must sit in a distinct frame with an explanation and must not be used as backgrounds.
+
+【Prohibited】Do not stack rounded cards without an information function; do not use glass effects, gradients, or luminous borders. Rounded containers are reserved for quotations, definitions, status, and process steps. Do not mix decorative colors; retain only one primary accent color per slide. Do not use full-slide hand-drawn textures or substitute strike-through marks for argumentation. Photographs/screenshots/illustrations may support evidence or a special cover, but must not become the default background. Do not center long passages of body text. Full-slide wine-red, black, dark-green, and cream reskinning must not jump between colors within the same chapter.
+
+【Slide Types and Layouts】Argument–evidence two-column｜One conclusion + explanation/source｜Left 40–45% text, right 55–60% chart/screenshot｜2–4 blocks｜Keyword color, quotation box, source line｜Not suitable for multiple equally weighted conclusions. Process/closed-loop diagram｜Method, lifecycle, causal chain｜Central 60–70% graphic with explanations on both sides｜1 main diagram, 4–10 nodes｜Semantic arrows and role glyphs｜Not suitable for a pure literature list. High-density data/definition slide｜Table + equation or multiple metrics｜35–45% table, 55–65% chart/definition box｜1–2 main charts or 1 table + 1 chart｜Fine horizontal rules, single-series highlight｜Not suitable for numbers without measurement definitions. Large-type viewpoint/chapter slide (special)｜Transition/strong judgment｜Full-slide 80–90% text, 0–1 metadata item｜1 statement｜Solid color or whitespace｜Not suitable for body evidence. Contents/navigation slide (special)｜Section transition｜Numbered list covering 60–80%, 3–5 items｜Current-item brush stroke/capsule｜Not suitable for chart explanation.
+
+【Density Baseline】There is no uniform minimum across the full deck. Two-column content slides typically contain 2–4 evidence blocks, 0–2 charts, and 1–3 explanation blocks, each approximately 20–60 words/characters equivalent; place whitespace between the title and evidence group and at the two-column divide. Process slides use 1 main diagram, 4–10 nodes, and 2–6 arrows. Table slides use 1 main table + 1 equation/conclusion box, with numbers right-aligned, text left-aligned, and 1–2 footnote lines; multi-chart slides use 2–4 small charts, label only key values, and do not crowd every bar with labels. Viewpoint slides contain only 1 statement of approximately 20–60 words/characters equivalent, while source notes still sit at the bottom. Close the “title—evidence” loop first, then move remaining copy down into the source line.