[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7138TCNDXCVVS5KC74J4B9C":3,"skill-same:01M7138TCNDXCVVS5KC74J4B9C":101},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":34,"publishedAt":40,"linkOnly":41,"markdownContent":42,"markdownBody":43,"frontMatter":44,"refName":45,"commitSha":46,"skillDirectory":47,"forksCount":48,"repositoryHtmlUrl":49,"repositoryHomepageUrl":6,"ownerHtmlUrl":50,"ownerBlogUrl":51,"ownerDescription":52,"updatedAt":53,"install":54,"resources":67},"01M7138TCNDXCVVS5KC74J4B9C","html-design-polish",null,"网页需要产品清晰度、信息层级、响应式或视觉系统美化、重设计或设计审计时使用。","EthanYoQ\u002FSkill-hub","EthanYoQ","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F257484394?v=4","01M70C9QKWR1YS0S7V26DNB18H","Ethan","01M70C9QMER047R5ANZZA6KC2P",11,"MIT","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002FSKILL.md",[18,23,26,29],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000009","前端开发","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":30,"categoryName":31,"taxonomyType":32,"taxonomyTypeDesc":33},"01JBSEEDCAT000000000000032","设计师","PROFESSION","适用职业",[35,36,37,38,39],"design","polish","html","html design polish","美化","2026-11-10 10:12:30",false,"---\nname: html-design-polish\ndescription: 网页需要产品清晰度、信息层级、响应式或视觉系统美化、重设计或设计审计时使用。\n---\n\n# HTML设计美化\n\n## Goal\n\nMake the product purpose and primary user task clear before adding visual character. Build quality through hierarchy, proportion, spacing, material, and brand fit; do not reproduce a fixed palette, font, radius, layout, or component library.\n\n## Scope and boundaries\n\nUse for product pages, internal tools, navigation hubs, mobile web tools, design audits, redesigns, visual refinement, reference-image interpretation, and pages that feel generic, templated, or brand-incoherent.\n\nDo not trigger by default for:\n\n- Ordinary feature implementation with an already-decided design.\n- A single CSS bug or mechanical responsive fix.\n- Pure code refactoring.\n- Exact pixel replication.\n- Component work with no product, hierarchy, or visual judgment.\n- Business research or content strategy without a web artifact.\n\nIn an existing project, preserve data, formulas, links, routes, states, business rules, and working behavior unless the brief explicitly changes them. Do not turn a focused tool into a landing page or dashboard merely to make it look more designed.\n\n## Choose the execution mode\n\n### Audit mode\n\nUse when the user asks to analyze, assess, audit, evaluate, recommend, or explicitly says not to modify yet.\n\n- Read the relevant project material and inspect the rendered page when possible.\n- Produce a diagnosis at the requested depth.\n- Do not edit code, generate implementation artifacts, or claim visual verification for changes that were not made.\n\n### Execute mode\n\nUse when the user asks to modify, redesign, optimize, implement, repair the design, or polish the page.\n\n1. Diagnose internally.\n2. Lock protected behavior and content.\n3. Implement high-confidence, in-scope changes directly.\n4. Run the page and verify rendered and business states.\n5. Complete a second refinement pass.\n6. Report at the task-appropriate detail level.\n\nDo not stop after routine diagnosis to request approval. An unfamiliar project is not a blocker: inspect it, establish invariants, and proceed.\n\nStop only if a critical input is absent, the project cannot be inspected or run enough to act safely, the next action creates irreversible\u002Fmaterial external risk, or the user must choose between materially different product directions.\n\n## Operating sequence\n\nUse this order. Do not start with colors, gradients, glass, or animation when an earlier layer fails.\n\n1. Product model\n2. User and core task\n3. Information architecture\n4. Interaction and responsive composition\n5. Visual system\n6. Browser and business verification\n\n### 1. Product model\n\nClassify the page before judging layout:\n\n- Single-task tool\n- Multi-step tool\n- Dashboard\n- Marketing landing page\n- Navigation hub\n- Content experience\n- Other explicit model\n\nConfirm that the current page model matches the actual product purpose. Diagnose these errors first:\n\n- A single-task tool is made into a dashboard.\n- An internal tool is made into a marketing landing page.\n- A result-oriented tool leads with extensive explanation.\n- A navigation hub is made into corporate promotion.\n- Unsupported metrics, cases, testimonials, or social proof are invented.\n- A reference image changes the product purpose instead of informing visual direction.\n\nIf the model is wrong, repair model and task sequence before visual polish.\n\n### 2. User and core task\n\nState, in one sentence each:\n\n- Primary user and use context.\n- Core task and desired outcome.\n- What a first-time user must understand in one viewport.\n- Primary action or result.\n- Secondary tasks and their frequency.\n- Protected data, formulas, links, routes, states, and business logic.\n- Target viewports and relevant device orientation.\n- Explicit scope and out-of-scope changes.\n\nAsk:\n\n- Can a new user say what this product is, who it serves, and what to do next?\n- Is the highest-value outcome visible before internal taxonomy, long explanation, or secondary reporting?\n- Does copy describe a concrete product task rather than generic slogans?\n- Is status\u002Favailability truthful?\n\nIf these answers are unclear, solve this layer before styling.\n\n### 3. Information architecture\n\nList visible modules in order. Rank each by user value and frequency.\n\nCheck:\n\n- Does the current order follow the user task rather than internal project structure?\n- Does the first viewport jointly expose purpose, current context, outcome, and next action?\n- Are primary, secondary, and supporting actions visibly different in weight?\n- Does any selector, navigation, or dashboard occupy more space than the content it controls?\n- Is a high-value result hidden behind long policy\u002Fprocess text?\n- Can low-frequency details be grouped, collapsed, deferred, or hidden?\n\nWhen several entries exist, group by user task and status. Do not flatten internal project names into equal cards. Promote the primary available action or result; make secondary entries quieter but discoverable.\n\n### 4. Interaction and responsive composition\n\nTreat mobile as a new composition and disclosure problem, not a reduced desktop layout.\n\nFor every responsive state decide:\n\n- What remains visible first.\n- What moves, collapses, or becomes on demand.\n- How navigation\u002Fselection changes: row, segmented control, compact grid, list, drawer, or another explicit pattern.\n- How long labels, values, units, and mixed-language text wrap.\n- How selected, focus-visible, expanded, disabled, loading, error, empty, and long-content states render when present.\n\nTest the actual minimum supported width, 375 px, 414 px, relevant tablet, and relevant landscape. Check for horizontal overflow, clipped controls, vertical character stacking, accidental multi-line clickable labels, loss of primary-task exposure, and touch-target failures.\n\nSelect columns from item count, task frequency, and label length. For example, a compact 3×2 grid can suit six navigation entries; it is not a global mobile rule.\n\n### 5. Visual system\n\nUse visual treatment to clarify purpose, hierarchy, state, and brand. Do not use it to hide structural weakness.\n\nInspect in this order when the page feels ordinary, heavy, generic, or “not premium”:\n\n1. Task order and first viewport.\n2. Container width and text measure.\n3. Alignment and spacing rhythm.\n4. Typography levels, numbers, labels, and line breaks.\n5. Density and visual weight.\n6. Asset quality and image crop.\n7. Borders, cards, shadows, material, and motion.\n\nGive every color, surface, border, icon, image, and animation a job: identity, hierarchy, state, grouping, affordance, depth, feedback, or explanation. Remove it if it has none.\n\nStandardize repeated geometry and interaction behavior. Preserve meaningful product\u002Fbrand differences such as logos and restrained product accents. “Unified” does not mean every item has identical color or prominence.\n\nUse a supplied reference image to extract hierarchy, proportion, material, contrast, image role, and interaction tone. Adapt those decisions to the product; do not copy pixels or change product purpose. Keep critical text in HTML rather than generated imagery.\n\n### Project icon generation (required)\n\nFor every named project, product, or service entry shown on the page, generate one corresponding primary icon with Image2 (`image_gen`). Do not substitute generic line icons, emoji, or a repeated placeholder as the primary project identifier.\n\n- Generate each icon from its actual product\u002Fservice function, name, and brand cues; use one distinct prompt per entry.\n- Use Apple TV-style app-icon language: clean rounded-square silhouette, tactile dimensional form, centered recognizable symbol, controlled highlight\u002Fshadow, no black outer ring, no nested frames, no baked UI chrome.\n- Keep the icon set structurally consistent: shared canvas, scale, corner treatment, depth, and lighting direction. Use restrained per-project color variation to aid recognition.\n- Do not bake important text, product names, or tiny labels into icons. Keep names and status as HTML beside or below the asset.\n- Inspect every generated result at its real card size. Regenerate individual failures instead of applying CSS effects to hide weak icon anatomy.\n- Save approved assets inside the current project and reference local copies. Never leave a project-used icon only in generated-image storage.\n- If the project already has an approved brand icon\u002Flogo, use it as the semantic reference; generate an Apple TV-style companion only when the user asked for the visual treatment or the existing asset cannot serve the entry.\n\n## Decision rules\n\n| Condition | Judge | Act |\n|---|---|---|\n| First viewport is unclear | Purpose, outcome, and next action are not visible together | Reorder\u002Frewrite the first viewport before styling |\n| A result tool starts with explanation | The result is delayed by low-frequency detail | Promote the result; collapse or defer explanation |\n| Users choose among many entries | Frequency, grouping, status, and scan order | Group by task; give primary\u002Favailable\u002Fsecondary entries distinct weight |\n| A selector dominates the page | Control uses more attention\u002Fspace than controlled content | Replace default-expanded cards with compact choices; expand only where needed |\n| The UI feels heavy | Nested frames, repeated labels, oversized cards, too many borders | Remove one structural layer at a time; retain only layers with a job |\n| The page feels generic | Product could be swapped by changing logo\u002Faccent only | Increase product-specific task language, structure, and asset role |\n| The page feels “not premium” | Hierarchy, proportion, spacing, type, density, or asset quality is weak | Correct those first; add effects only if they strengthen the direction |\n| Strong material\u002Fstyle is requested | It supports brand but risks legibility | Build controlled layers; protect contrast, focus, and reduced motion |\n| Reference image is supplied | DNA is useful but product differs | Adapt visual DNA, retain current product task and invariants |\n| Visual-only change is requested | Behavior must remain unchanged | Lock invariants and exercise them before\u002Fafter |\n| Brand variants coexist | Uniformity may erase recognition | Standardize system rules; retain meaningful identity signals |\n| Mobile is crowded | Desktop order\u002Fdensity no longer serves the task | Reorder, regroup, collapse, resize, and retest |\n\n## Avoid these failure modes\n\n- Effects, gradients, glass, or animation used to compensate for weak hierarchy.\n- Nested frames around already distinctive logos or assets.\n- Large fields of same-size, same-weight cards.\n- Every section given an icon, pill, badge, gradient, and rounded container.\n- Hero copy that sounds polished but does not explain the tool.\n- A hidden or diluted primary action.\n- Generic SaaS rhythm imposed on a focused workflow.\n- Dashboard chrome imposed on an internal single-task tool.\n- Desktop layout merely shrunk for mobile.\n- Generated or invented metrics, testimonials, proof, or case studies.\n- Important text embedded in generated images.\n- Visual changes that silently alter formulas, policy data, links, state, or copy meaning.\n- Declaring success from source inspection or one default screenshot alone.\n\n## Implementation workflow for Execute mode\n\n### Establish invariants\n\nBefore editing, identify and protect:\n\n- Business data, formulas, and result calculations.\n- URLs, routes, product names, availability, and feature flags.\n- Existing interactions and keyboard behavior.\n- Accessibility requirements and reduced-motion behavior.\n- Required root\u002Fproduction parity, if applicable.\n- Explicitly out-of-scope modules or files.\n\nCreate or run the smallest useful checks for these invariants. A visual improvement that breaks a business rule is a regression.\n\n### Make structural changes first\n\nChange page order, grouping, disclosure, navigation, or CTA weight only when product model, task sequence, or hierarchy requires it. Otherwise patch locally.\n\nAfter structure changes, verify protected behavior before visual polish. If the task is still unclear with colors\u002Feffects mentally removed, continue structural work.\n\n### Build the system\n\nReuse sound existing tokens. When creating new rules, define named roles for typography, spacing, surface, border, focus, status, motion, and brand accents. Avoid per-element visual improvisation.\n\nUse containers for repeated entities, state, interaction, or meaningful grouping. Use icons for navigation, status, or rapid recognition; do not attach them automatically to every heading or row.\n\nFor project\u002Fservice entry cards, run the required Image2 icon-generation pass before final card polish. Generate first, place real assets, then tune crop, vessel, contrast, and spacing around the approved icons. Do not judge the card system from placeholder icons.\n\n### Implement responsive behavior\n\nDesign mobile deliberately:\n\n1. Choose the first-visible information.\n2. Set the mobile order.\n3. Choose collapsed\u002Fexpanded defaults.\n4. Select a compact navigation or selector form.\n5. Add safe wrapping and minimum-width behavior.\n6. Test long labels, values, units, and mixed-language strings.\n7. Add landscape-height rules when the usage context requires them.\n\n### Verify and refine\n\nRun the actual page. At each target viewport, record:\n\n- Whether the core task is visible and understandable.\n- Horizontal overflow and clipping.\n- Typography, wrapping, alignment, and image crop.\n- Selected, focus, expanded, long-content, error\u002Floading\u002Fempty states when present.\n- Console errors and asset loading.\n- Relevant business outputs and interactions.\n\nCompare the same state before and after. Then perform a second refinement pass for line breaks, spacing, content width, same-weight residue, focus\u002Fselected treatment, numeric alignment, image contrast, and reduced motion.\n\n## Audit template\n\nUse this structure in Audit mode or as internal notes in Execute mode.\n\n### Context\n\n- Product model:\n- Primary user\u002Fcontext:\n- Core task and first-viewport outcome:\n- Protected behavior:\n- Target states\u002Fviewports:\n- Scope and exclusions:\n\n### Ranked diagnosis\n\n1. Product model:\n2. User and core task:\n3. Information architecture:\n4. Interaction and responsive composition:\n5. Visual system and brand fit:\n6. Browser\u002Fbusiness verification gaps:\n\n### Direction\n\n- What should become dominant:\n- What should become quieter, grouped, collapsed, or deferred:\n- Brand\u002Freference decisions to retain:\n- Effects deliberately excluded:\n- Invariants to protect:\n- Evidence required to accept the result:\n\n## Acceptance checklist\n\n- [ ] Product model matches product purpose.\n- [ ] User, core task, and first-viewport outcome are explicit.\n- [ ] Product\u002FIA issues were resolved before visual polish.\n- [ ] Primary and secondary actions have distinct weight.\n- [ ] Protected behavior and content invariants still pass.\n- [ ] Visual choices serve task, brand, state, or reference evidence.\n- [ ] Each visible project\u002Fservice entry has an approved, project-specific Image2 icon in the Apple TV-style family.\n- [ ] No unnecessary decorative nesting or equal-card field remains.\n- [ ] Mobile was recomposed where needed, not only scaled down.\n- [ ] Target desktop\u002Fmobile\u002Ftablet\u002Flandscape states were rendered.\n- [ ] Overflow, wrapping, focus, selected, collapsed\u002Fexpanded, and relevant business states were checked.\n- [ ] Before\u002Fafter comparison proves clearer task use, not only different styling.\n- [ ] A second refinement pass is complete.\n\n## Output contract\n\nChoose output depth from task scale. Do not create a full design report for a small targeted refinement.\n\n### Small targeted refinement\n\nReturn only:\n\n1. Core issue\n2. Changes made\n3. Browser verification\n4. Remaining issues\n\n### Full redesign or audit\n\nReturn, in order:\n\n1. Product and task summary\n2. Ranked diagnosis: product model, information architecture, interaction\u002Fresponsive, visual system\n3. Protected invariants and project-specific constraints\n4. Chosen direction and rejected alternatives with reasons\n5. Files and states changed, or proposed changes in Audit mode\n6. Browser verification matrix and before\u002Fafter findings\n7. Remaining risks or evidence gaps\n\nNever present a contextual palette, font, radius, card form, or layout as a universal rule.\n","# HTML设计美化\n\n## Goal\n\nMake the product purpose and primary user task clear before adding visual character. Build quality through hierarchy, proportion, spacing, material, and brand fit; do not reproduce a fixed palette, font, radius, layout, or component library.\n\n## Scope and boundaries\n\nUse for product pages, internal tools, navigation hubs, mobile web tools, design audits, redesigns, visual refinement, reference-image interpretation, and pages that feel generic, templated, or brand-incoherent.\n\nDo not trigger by default for:\n\n- Ordinary feature implementation with an already-decided design.\n- A single CSS bug or mechanical responsive fix.\n- Pure code refactoring.\n- Exact pixel replication.\n- Component work with no product, hierarchy, or visual judgment.\n- Business research or content strategy without a web artifact.\n\nIn an existing project, preserve data, formulas, links, routes, states, business rules, and working behavior unless the brief explicitly changes them. Do not turn a focused tool into a landing page or dashboard merely to make it look more designed.\n\n## Choose the execution mode\n\n### Audit mode\n\nUse when the user asks to analyze, assess, audit, evaluate, recommend, or explicitly says not to modify yet.\n\n- Read the relevant project material and inspect the rendered page when possible.\n- Produce a diagnosis at the requested depth.\n- Do not edit code, generate implementation artifacts, or claim visual verification for changes that were not made.\n\n### Execute mode\n\nUse when the user asks to modify, redesign, optimize, implement, repair the design, or polish the page.\n\n1. Diagnose internally.\n2. Lock protected behavior and content.\n3. Implement high-confidence, in-scope changes directly.\n4. Run the page and verify rendered and business states.\n5. Complete a second refinement pass.\n6. Report at the task-appropriate detail level.\n\nDo not stop after routine diagnosis to request approval. An unfamiliar project is not a blocker: inspect it, establish invariants, and proceed.\n\nStop only if a critical input is absent, the project cannot be inspected or run enough to act safely, the next action creates irreversible\u002Fmaterial external risk, or the user must choose between materially different product directions.\n\n## Operating sequence\n\nUse this order. Do not start with colors, gradients, glass, or animation when an earlier layer fails.\n\n1. Product model\n2. User and core task\n3. Information architecture\n4. Interaction and responsive composition\n5. Visual system\n6. Browser and business verification\n\n### 1. Product model\n\nClassify the page before judging layout:\n\n- Single-task tool\n- Multi-step tool\n- Dashboard\n- Marketing landing page\n- Navigation hub\n- Content experience\n- Other explicit model\n\nConfirm that the current page model matches the actual product purpose. Diagnose these errors first:\n\n- A single-task tool is made into a dashboard.\n- An internal tool is made into a marketing landing page.\n- A result-oriented tool leads with extensive explanation.\n- A navigation hub is made into corporate promotion.\n- Unsupported metrics, cases, testimonials, or social proof are invented.\n- A reference image changes the product purpose instead of informing visual direction.\n\nIf the model is wrong, repair model and task sequence before visual polish.\n\n### 2. User and core task\n\nState, in one sentence each:\n\n- Primary user and use context.\n- Core task and desired outcome.\n- What a first-time user must understand in one viewport.\n- Primary action or result.\n- Secondary tasks and their frequency.\n- Protected data, formulas, links, routes, states, and business logic.\n- Target viewports and relevant device orientation.\n- Explicit scope and out-of-scope changes.\n\nAsk:\n\n- Can a new user say what this product is, who it serves, and what to do next?\n- Is the highest-value outcome visible before internal taxonomy, long explanation, or secondary reporting?\n- Does copy describe a concrete product task rather than generic slogans?\n- Is status\u002Favailability truthful?\n\nIf these answers are unclear, solve this layer before styling.\n\n### 3. Information architecture\n\nList visible modules in order. Rank each by user value and frequency.\n\nCheck:\n\n- Does the current order follow the user task rather than internal project structure?\n- Does the first viewport jointly expose purpose, current context, outcome, and next action?\n- Are primary, secondary, and supporting actions visibly different in weight?\n- Does any selector, navigation, or dashboard occupy more space than the content it controls?\n- Is a high-value result hidden behind long policy\u002Fprocess text?\n- Can low-frequency details be grouped, collapsed, deferred, or hidden?\n\nWhen several entries exist, group by user task and status. Do not flatten internal project names into equal cards. Promote the primary available action or result; make secondary entries quieter but discoverable.\n\n### 4. Interaction and responsive composition\n\nTreat mobile as a new composition and disclosure problem, not a reduced desktop layout.\n\nFor every responsive state decide:\n\n- What remains visible first.\n- What moves, collapses, or becomes on demand.\n- How navigation\u002Fselection changes: row, segmented control, compact grid, list, drawer, or another explicit pattern.\n- How long labels, values, units, and mixed-language text wrap.\n- How selected, focus-visible, expanded, disabled, loading, error, empty, and long-content states render when present.\n\nTest the actual minimum supported width, 375 px, 414 px, relevant tablet, and relevant landscape. Check for horizontal overflow, clipped controls, vertical character stacking, accidental multi-line clickable labels, loss of primary-task exposure, and touch-target failures.\n\nSelect columns from item count, task frequency, and label length. For example, a compact 3×2 grid can suit six navigation entries; it is not a global mobile rule.\n\n### 5. Visual system\n\nUse visual treatment to clarify purpose, hierarchy, state, and brand. Do not use it to hide structural weakness.\n\nInspect in this order when the page feels ordinary, heavy, generic, or “not premium”:\n\n1. Task order and first viewport.\n2. Container width and text measure.\n3. Alignment and spacing rhythm.\n4. Typography levels, numbers, labels, and line breaks.\n5. Density and visual weight.\n6. Asset quality and image crop.\n7. Borders, cards, shadows, material, and motion.\n\nGive every color, surface, border, icon, image, and animation a job: identity, hierarchy, state, grouping, affordance, depth, feedback, or explanation. Remove it if it has none.\n\nStandardize repeated geometry and interaction behavior. Preserve meaningful product\u002Fbrand differences such as logos and restrained product accents. “Unified” does not mean every item has identical color or prominence.\n\nUse a supplied reference image to extract hierarchy, proportion, material, contrast, image role, and interaction tone. Adapt those decisions to the product; do not copy pixels or change product purpose. Keep critical text in HTML rather than generated imagery.\n\n### Project icon generation (required)\n\nFor every named project, product, or service entry shown on the page, generate one corresponding primary icon with Image2 (`image_gen`). Do not substitute generic line icons, emoji, or a repeated placeholder as the primary project identifier.\n\n- Generate each icon from its actual product\u002Fservice function, name, and brand cues; use one distinct prompt per entry.\n- Use Apple TV-style app-icon language: clean rounded-square silhouette, tactile dimensional form, centered recognizable symbol, controlled highlight\u002Fshadow, no black outer ring, no nested frames, no baked UI chrome.\n- Keep the icon set structurally consistent: shared canvas, scale, corner treatment, depth, and lighting direction. Use restrained per-project color variation to aid recognition.\n- Do not bake important text, product names, or tiny labels into icons. Keep names and status as HTML beside or below the asset.\n- Inspect every generated result at its real card size. Regenerate individual failures instead of applying CSS effects to hide weak icon anatomy.\n- Save approved assets inside the current project and reference local copies. Never leave a project-used icon only in generated-image storage.\n- If the project already has an approved brand icon\u002Flogo, use it as the semantic reference; generate an Apple TV-style companion only when the user asked for the visual treatment or the existing asset cannot serve the entry.\n\n## Decision rules\n\n| Condition | Judge | Act |\n|---|---|---|\n| First viewport is unclear | Purpose, outcome, and next action are not visible together | Reorder\u002Frewrite the first viewport before styling |\n| A result tool starts with explanation | The result is delayed by low-frequency detail | Promote the result; collapse or defer explanation |\n| Users choose among many entries | Frequency, grouping, status, and scan order | Group by task; give primary\u002Favailable\u002Fsecondary entries distinct weight |\n| A selector dominates the page | Control uses more attention\u002Fspace than controlled content | Replace default-expanded cards with compact choices; expand only where needed |\n| The UI feels heavy | Nested frames, repeated labels, oversized cards, too many borders | Remove one structural layer at a time; retain only layers with a job |\n| The page feels generic | Product could be swapped by changing logo\u002Faccent only | Increase product-specific task language, structure, and asset role |\n| The page feels “not premium” | Hierarchy, proportion, spacing, type, density, or asset quality is weak | Correct those first; add effects only if they strengthen the direction |\n| Strong material\u002Fstyle is requested | It supports brand but risks legibility | Build controlled layers; protect contrast, focus, and reduced motion |\n| Reference image is supplied | DNA is useful but product differs | Adapt visual DNA, retain current product task and invariants |\n| Visual-only change is requested | Behavior must remain unchanged | Lock invariants and exercise them before\u002Fafter |\n| Brand variants coexist | Uniformity may erase recognition | Standardize system rules; retain meaningful identity signals |\n| Mobile is crowded | Desktop order\u002Fdensity no longer serves the task | Reorder, regroup, collapse, resize, and retest |\n\n## Avoid these failure modes\n\n- Effects, gradients, glass, or animation used to compensate for weak hierarchy.\n- Nested frames around already distinctive logos or assets.\n- Large fields of same-size, same-weight cards.\n- Every section given an icon, pill, badge, gradient, and rounded container.\n- Hero copy that sounds polished but does not explain the tool.\n- A hidden or diluted primary action.\n- Generic SaaS rhythm imposed on a focused workflow.\n- Dashboard chrome imposed on an internal single-task tool.\n- Desktop layout merely shrunk for mobile.\n- Generated or invented metrics, testimonials, proof, or case studies.\n- Important text embedded in generated images.\n- Visual changes that silently alter formulas, policy data, links, state, or copy meaning.\n- Declaring success from source inspection or one default screenshot alone.\n\n## Implementation workflow for Execute mode\n\n### Establish invariants\n\nBefore editing, identify and protect:\n\n- Business data, formulas, and result calculations.\n- URLs, routes, product names, availability, and feature flags.\n- Existing interactions and keyboard behavior.\n- Accessibility requirements and reduced-motion behavior.\n- Required root\u002Fproduction parity, if applicable.\n- Explicitly out-of-scope modules or files.\n\nCreate or run the smallest useful checks for these invariants. A visual improvement that breaks a business rule is a regression.\n\n### Make structural changes first\n\nChange page order, grouping, disclosure, navigation, or CTA weight only when product model, task sequence, or hierarchy requires it. Otherwise patch locally.\n\nAfter structure changes, verify protected behavior before visual polish. If the task is still unclear with colors\u002Feffects mentally removed, continue structural work.\n\n### Build the system\n\nReuse sound existing tokens. When creating new rules, define named roles for typography, spacing, surface, border, focus, status, motion, and brand accents. Avoid per-element visual improvisation.\n\nUse containers for repeated entities, state, interaction, or meaningful grouping. Use icons for navigation, status, or rapid recognition; do not attach them automatically to every heading or row.\n\nFor project\u002Fservice entry cards, run the required Image2 icon-generation pass before final card polish. Generate first, place real assets, then tune crop, vessel, contrast, and spacing around the approved icons. Do not judge the card system from placeholder icons.\n\n### Implement responsive behavior\n\nDesign mobile deliberately:\n\n1. Choose the first-visible information.\n2. Set the mobile order.\n3. Choose collapsed\u002Fexpanded defaults.\n4. Select a compact navigation or selector form.\n5. Add safe wrapping and minimum-width behavior.\n6. Test long labels, values, units, and mixed-language strings.\n7. Add landscape-height rules when the usage context requires them.\n\n### Verify and refine\n\nRun the actual page. At each target viewport, record:\n\n- Whether the core task is visible and understandable.\n- Horizontal overflow and clipping.\n- Typography, wrapping, alignment, and image crop.\n- Selected, focus, expanded, long-content, error\u002Floading\u002Fempty states when present.\n- Console errors and asset loading.\n- Relevant business outputs and interactions.\n\nCompare the same state before and after. Then perform a second refinement pass for line breaks, spacing, content width, same-weight residue, focus\u002Fselected treatment, numeric alignment, image contrast, and reduced motion.\n\n## Audit template\n\nUse this structure in Audit mode or as internal notes in Execute mode.\n\n### Context\n\n- Product model:\n- Primary user\u002Fcontext:\n- Core task and first-viewport outcome:\n- Protected behavior:\n- Target states\u002Fviewports:\n- Scope and exclusions:\n\n### Ranked diagnosis\n\n1. Product model:\n2. User and core task:\n3. Information architecture:\n4. Interaction and responsive composition:\n5. Visual system and brand fit:\n6. Browser\u002Fbusiness verification gaps:\n\n### Direction\n\n- What should become dominant:\n- What should become quieter, grouped, collapsed, or deferred:\n- Brand\u002Freference decisions to retain:\n- Effects deliberately excluded:\n- Invariants to protect:\n- Evidence required to accept the result:\n\n## Acceptance checklist\n\n- [ ] Product model matches product purpose.\n- [ ] User, core task, and first-viewport outcome are explicit.\n- [ ] Product\u002FIA issues were resolved before visual polish.\n- [ ] Primary and secondary actions have distinct weight.\n- [ ] Protected behavior and content invariants still pass.\n- [ ] Visual choices serve task, brand, state, or reference evidence.\n- [ ] Each visible project\u002Fservice entry has an approved, project-specific Image2 icon in the Apple TV-style family.\n- [ ] No unnecessary decorative nesting or equal-card field remains.\n- [ ] Mobile was recomposed where needed, not only scaled down.\n- [ ] Target desktop\u002Fmobile\u002Ftablet\u002Flandscape states were rendered.\n- [ ] Overflow, wrapping, focus, selected, collapsed\u002Fexpanded, and relevant business states were checked.\n- [ ] Before\u002Fafter comparison proves clearer task use, not only different styling.\n- [ ] A second refinement pass is complete.\n\n## Output contract\n\nChoose output depth from task scale. Do not create a full design report for a small targeted refinement.\n\n### Small targeted refinement\n\nReturn only:\n\n1. Core issue\n2. Changes made\n3. Browser verification\n4. Remaining issues\n\n### Full redesign or audit\n\nReturn, in order:\n\n1. Product and task summary\n2. Ranked diagnosis: product model, information architecture, interaction\u002Fresponsive, visual system\n3. Protected invariants and project-specific constraints\n4. Chosen direction and rejected alternatives with reasons\n5. Files and states changed, or proposed changes in Audit mode\n6. Browser verification matrix and before\u002Fafter findings\n7. Remaining risks or evidence gaps\n\nNever present a contextual palette, font, radius, card form, or layout as a universal rule.\n",{"name":5,"description":7},"mine","1e7390741dffd3d5041fca4ebc056071ee3ba137","skills\u002F07-media-content\u002Fhtml-design-polish",5,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub","https:\u002F\u002Fgithub.com\u002FEthanYoQ","www.orz.md","Part Marketer, Part AI Explorer.","2026-11-07 14:52:29",{"source":8,"skillName":5,"commands":55,"prompt":65,"zipAvailable":66},[56,59,62],{"tool":57,"command":58},"npx","npx skills add EthanYoQ\u002FSkill-hub --skill html-design-polish",{"tool":60,"command":61},"bunx","bunx skills add EthanYoQ\u002FSkill-hub --skill html-design-polish",{"tool":63,"command":64},"pnpm","pnpm dlx skills add EthanYoQ\u002FSkill-hub --skill html-design-polish","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002FSKILL.md",true,[68,74,78,82,88,94,97],{"path":69,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":72,"sourceHtmlUrl":73},"assets\u002Fclovernas-navigation-reference.png","ASSET","资产文件",2943405,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002Fassets\u002Fclovernas-navigation-reference.png",{"path":75,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":76,"sourceHtmlUrl":77},"assets\u002Finnovation-navigation-reference.png",7153296,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002Fassets\u002Finnovation-navigation-reference.png",{"path":79,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":80,"sourceHtmlUrl":81},"assets\u002Fxacduro-navigation-reference.png",6108455,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002Fassets\u002Fxacduro-navigation-reference.png",{"path":83,"resourceType":84,"resourceTypeDesc":85,"sizeBytes":86,"sourceHtmlUrl":87},"CHANGELOG.md","REFERENCE","参考文档",323,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002FCHANGELOG.md",{"path":89,"resourceType":90,"resourceTypeDesc":91,"sizeBytes":92,"sourceHtmlUrl":93},"docs\u002F.gitkeep","OTHER","其他",0,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002Fdocs\u002F.gitkeep",{"path":95,"resourceType":90,"resourceTypeDesc":91,"sizeBytes":92,"sourceHtmlUrl":96},"examples\u002F.gitkeep","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002Fexamples\u002F.gitkeep",{"path":98,"resourceType":84,"resourceTypeDesc":85,"sizeBytes":99,"sourceHtmlUrl":100},"README.md",2159,"https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F07-media-content\u002Fhtml-design-polish\u002FREADME.md",[102,130,159,183,201,216],{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":109,"publishedAt":40},"01M713CBDSYARG4PXFXK2R2NNN","xhs-pharma-social-listening","Run evidence-disciplined Xiaohongshu social listening for foreign pharma employee pain points. Use when the user asks to collect and analyze notes\u002Fcomments about pharma companies, aliases, roles, compliance, medical affairs, or AI topics, then produce a cited and scored sampled-insight report. Includes independent-evidence counting, MediaCrawler batch collection, and CAPTCHA\u002F461 fallback to Agent Reach\u002FOpenCLI.","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F10-business-industry\u002Fxhs-pharma-social-listening\u002FSKILL.md",[108],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129],"social","xhs pharma social listening","social listening","xhs","medical","counting","when","collection","batch","analyze","notes","run","opencli","fallback","captcha","agent reach","evidence","for","collect","produce",{"skillNo":131,"name":132,"nameZh":6,"description":133,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":134,"categories":135,"keywords":140,"publishedAt":40},"01M713C8Q76ZP0WXXBRAFKS6EE","healthcare-phi-compliance","Protected Health Information (PHI) and Personally Identifiable Information (PII) compliance patterns for healthcare applications. Covers data classification, access control, audit trails, encryption, and common leak vectors. Use when code touches PHI or PII in a healthcare system, or when auditing access control, audit trails, or leak vectors.","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F10-business-industry\u002Fhealthcare-phi-compliance\u002FSKILL.md",[136,137],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":138,"categoryName":139,"taxonomyType":32,"taxonomyTypeDesc":33},"01JBSEEDCAT000000000000037","律师与法务",[141,142,143,144,145,146,147,148,116,149,150,151,152,153,154,155,127,156,157,158],"healthcare","phi","compliance","healthcare phi compliance","pii","compliance patterns","information","access","encryption","code","common","audit","data","patterns","Access Control","system","applications","Auditing Access Control",{"skillNo":160,"name":161,"nameZh":6,"description":162,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":163,"categories":164,"keywords":166,"publishedAt":40},"01M713C5R2GE43X45T2K75TB76","healthcare-eval-harness","Patient safety evaluation harness for healthcare application deployments. Automated test suites for CDSS accuracy, PHI exposure, clinical workflow integrity, and integration compliance. Blocks deployments on safety failures. Use when a healthcare deployment must be gated on patient-safety tests for CDSS accuracy, PHI exposure, and workflow integrity.","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F10-business-industry\u002Fhealthcare-eval-harness\u002FSKILL.md",[165],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[141,167,168,169,170,171,172,116,173,174,175,176,142,127,177,178,179,180,181,182],"eval","harness","healthcare eval harness","eval harness","clinical","patient","workflow","Evaluation","Safety","integrity","deployment","accuracy","tests","deployments","use","super",{"skillNo":184,"name":185,"nameZh":6,"description":186,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":187,"categories":188,"keywords":190,"publishedAt":40},"01M713C2J17S91VYR7A8M696VR","healthcare-emr-patterns","EMR\u002FEHR development patterns for healthcare applications. Clinical safety, encounter workflows, prescription generation, clinical decision support integration, and accessibility-first UI for medical data entry. Use when building EMR or EHR features such as encounter workflows, prescription generation, or clinical data entry UI.","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F10-business-industry\u002Fhealthcare-emr-patterns\u002FSKILL.md",[189],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[141,191,154,192,193,114,171,194,116,195,196,197,175,153,198,199,200,127,157,181],"EMR","healthcare emr patterns","decision","workflows","support","clinical decision support","features","ui","as","development",{"skillNo":202,"name":203,"nameZh":6,"description":204,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":205,"categories":206,"keywords":208,"publishedAt":40},"01M713BZERTXV0N8X6SW9DDVWZ","healthcare-cdss-patterns","Clinical Decision Support System (CDSS) development patterns. Drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), alert severity classification, and integration into EMR workflows. Use when building clinical decision support — drug interaction checks, dose validation, clinical scoring, or alert severity.","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F10-business-industry\u002Fhealthcare-cdss-patterns\u002FSKILL.md",[207],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[141,154,209,193,171,191,194,116,210,211,195,196,212,200,156,213,181,214,182,215],"healthcare cdss patterns","scoring","interaction","severity","classification","alert","decision support",{"skillNo":217,"name":218,"nameZh":6,"description":219,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":220,"categories":221,"keywords":223,"publishedAt":40},"01M713BVK3PQR7V4ZKYGMWMXJD","dxy-task-clarify","用户的请求含糊、缺背景、缺资料说明、缺读者或用途时使用。先用少量带选项的追问把任务问清楚，写成任务卡，用户确认后再执行。尤其适用于医生、科研人员和药企各岗位的工作，例如「帮我分析一下这些数据」「帮我准备明天的拜访」「写个总结」「做个 PPT」「回一下这封邮件」「整理一下会议纪要」「整理一下这些文献」「帮我准备下周的讲课」「写个病例汇报」「帮我写个提示词」。用户说「先问我」「帮我理清需求」「我不知道怎么描述」「我不会问 AI」时也使用。用途、资料、读者都已说清的请求，纯翻译，以及写代码、改代码，不使用。","https:\u002F\u002Fgithub.com\u002FEthanYoQ\u002FSkill-hub\u002Fblob\u002Fmine\u002Fskills\u002F10-business-industry\u002Fdxy-task-clarify\u002FSKILL.md",[222],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[224,225,226,227,228,229,230,231],"clarify","dxy task clarify","task","pptx","整理","翻译","数据","总结"]