[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6CHTP6F4MENDBRDT7QQBKNY":3,"skill-same:01M6CHTP6F4MENDBRDT7QQBKNY":134},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":27,"publishedAt":49,"linkOnly":50,"markdownContent":51,"markdownBody":52,"frontMatter":53,"refName":54,"commitSha":55,"skillDirectory":56,"forksCount":57,"repositoryHtmlUrl":58,"repositoryHomepageUrl":59,"ownerHtmlUrl":60,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":61,"install":62,"resources":75},"01M6CHTP6F4MENDBRDT7QQBKNY","sero-humanize",null,"Use to draft, edit, or audit Sero application content, including UI copy,\ndocumentation, README text, and release notes. Use unslop for regular chat\nand status reports instead. Not for code review or creative and promotional\nwriting.","sero-labs\u002Fsero","sero-labs","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F278893820?v=4","01M6CHSF0GA1ZZSBGVB5ESH37Q","01M6CHSF4WD1EGGY0RNFHQ5R4N",28,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fsero-humanize\u002FSKILL.md",[17,22],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000007","代码审查","SCENARIO","使用场景",{"categoryNo":23,"categoryName":24,"taxonomyType":25,"taxonomyTypeDesc":26},"01JBSEEDCAT000000000000036","内容创作者","PROFESSION","适用职业",[28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48],"sero humanize","humanize","ui","for","reports","creative","use","edit","README","text","unslop","code review","audit","release","code","release notes","notes","documentation","status","review","writing","2026-11-10 10:12:30",false,"---\nname: sero-humanize\ndescription: |\n  Use to draft, edit, or audit Sero application content, including UI copy,\n  documentation, README text, and release notes. Use unslop for regular chat\n  and status reports instead. Not for code review or creative and promotional\n  writing.\n---\n\n# Sero Humanize\n\nMake Sero prose direct, specific, and useful. Human writing in technical\ndocumentation does not need personality. It needs clear decisions, concrete\nfacts, and respect for the reader's time.\n\nApply this skill to Sero content even when drafting it in chat. Use `unslop`\nfor the surrounding conversation, not for the content itself.\n\n## Follow the requested mode\n\n- For an audit or review, report the material patterns and do not edit files.\n- For an edit, rewrite the named files in place.\n- For a new document, apply these rules while drafting it.\n- If the request does not specify a mode, infer it from the requested action.\n  Do not turn a request to assess prose into permission to change it.\n\n## Establish the voice and preservation set\n\nBefore editing:\n\n1. For a passage edit, read the passage and enough surrounding text to preserve\n   meaning. Read the full file for a whole-document audit or structural rewrite.\n2. Read the nearest repository instructions that apply to those files.\n3. Use adjacent, clearly human-edited Sero documentation as the voice sample\n   when the named files do not establish a consistent voice.\n4. Record what must not change:\n   - technical meaning and product behaviour;\n   - product names, canonical terms, and exact UI labels;\n   - commands, code, file paths, numbers, limits, and factual claims;\n   - the documentation type, useful narrative flow, and balance between prose,\n     lists, tables, examples, and callouts;\n   - frontmatter, anchors, link targets, image paths, and screenshot order;\n   - the purpose and placement of each image, diagram, and other media asset;\n   - quotations and user input examples, unless the user asks to edit them.\n\nIdentify the intended reader. Unless the page states otherwise, assume the\nreader knows neither Sero nor the feature. Do not assume that simpler grammar\nfixes an explanation that requires missing product knowledge.\n\nDo not add a fact to make a sentence more vivid. Verify a doubtful claim from\nthe repository or leave it unchanged and report the doubt.\n\n## Audit structure before wording\n\nLook for clusters and repeated patterns. Do not treat one punctuation mark or\none common word as proof of AI writing.\n\nPrioritize these defects:\n\n- Meta narration that announces the next explanation instead of giving it.\n- Repeated tutorial staging such as \"what you are about to learn\" and \"what you\n  have learned.\"\n- Fixed enumerations such as \"three things are worth noticing\" when a direct\n  heading or short list is clearer.\n- Several paragraphs that can change order without changing the argument.\n- A heading followed by a sentence that only repeats the heading.\n- A conclusion that repeats the introduction without adding an action or fact.\n- Repeated summaries of the same screen, process, or result.\n- Forced contrasts such as \"not only X, but Y\" or \"not X; rather Y.\"\n- Groups of three used for rhythm instead of meaning.\n- Mechanical bold lead-ins, excessive inline bold, or lists that should be\n  short prose.\n- Promotional adjectives, vague importance claims, and unsupported praise.\n- Vague actors, passive constructions, filler, stacked hedges, and abstract\n  nouns where an action is available.\n- Synonym cycling for one product concept. Repeat the canonical term.\n- Long sentences that mix instructions, exceptions, and background.\n- Em dashes used repeatedly to join thoughts that need separate sentences.\n- Headings that narrate the demo or expose implementation language instead of\n  naming the reader's task, such as \"The finish\" or \"Answer the gate.\"\n- Examples that depend on an unexplained demo domain and therefore do not help\n  the reader understand the feature.\n- Tutorials that start using the product before they give prerequisites,\n  sample data, expected starting state, or required sign-ins.\n- Result sections that recite one captured run instead of telling the reader\n  what to inspect and verify.\n\nKeep useful structure. A list, summary, warning, question heading, or em dash is\nnot a defect by itself.\n\nDo not normalize a page or site to one format. Humanize the defective passages,\nnot every paragraph. If a page already explains a concept well in prose, keep\nit as prose.\n\n## Rewrite for Sero documentation\n\nApply ASD-STE100 Simplified Technical English where it fits the material:\n\n- Put the action or answer first.\n- Use active voice when the actor matters.\n- Give one main instruction per sentence.\n- Put a condition before the action when the reader must know it first.\n- Prefer common, precise words over formal or promotional alternatives.\n- Use the same term for the same thing.\n- Keep paragraphs focused on one subject.\n- Keep necessary limits, cautions, and exceptions close to the action.\n- Use contractions only when the established local voice requires them.\n- Keep exact UI text in bold when the documentation uses bold for controls.\n- Keep code identifiers and paths in code formatting.\n- Retain a summary only when it helps the reader decide or act.\n\nUse lists only when the content is naturally a sequence, set of choices,\nchecklist, or compact reference. Do not:\n\n- convert explanatory prose into bullet points only to make it shorter;\n- turn each sentence or paragraph into a list item;\n- replace transitions and reasoning with disconnected bullets;\n- use repeated lists where a short paragraph gives the reader necessary\n  context; or\n- make several pages share the same mechanical list structure.\n\nAfter the sentence pass, read the page as a whole. If lists now dominate a page\nthat previously used useful prose, restore the prose. Clear technical writing\nneeds connected explanation as well as scannable reference material.\n\nFor an overview page:\n\n- Explain the feature in familiar words before using its product terms.\n- State what the user gives Sero, what Sero does, and what the user reviews.\n- When comparing features, give one plain decision rule. Use examples that a\n  reader can understand without knowing the tutorial repository or a specialist\n  software domain.\n\nFor a tutorial:\n\n- Put setup before the first product action. Include required software,\n  accounts, sign-ins, repository or sample-data setup, and a command or visible\n  result that confirms the expected starting state.\n- Prefer a stable sample repository over instructions that ask an agent to\n  generate approximate sample data. Verify the repository contents and commands\n  before documenting them.\n- Use task-based headings such as \"Review the plan,\" \"Change the plan,\" and\n  \"Check the result.\" A heading must describe the full purpose of its section;\n  do not narrow a general control to one example case.\n- End with checks the reader can perform. Do not use a captured run's cost,\n  duration, names, or outcome as a substitute for verification instructions.\n\nFor feature language:\n\n- Use the visible object name: icon, button, tab, question, or approval request.\n  Do not call an icon a mark or expose internal terms such as gate, fan-out, or\n  feedback route when plain behaviour is enough.\n- Keep exact UI labels unchanged, but explain them with common words.\n- Put high-value quality-of-life features where readers will find them. Give\n  them enough space to explain when the control appears, how to use it, what it\n  changes, and what remains under user control.\n\nCompress or merge only the passages that contain a verified structural defect.\nKeep useful depth, examples, transitions, and paragraph structure. A shorter\npage is not automatically a better page.\n\nDo not manufacture a human voice with:\n\n- anecdotes, opinions, jokes, sensory details, or personal asides;\n- fragments, one-word sentences, or dramatic punch lines;\n- arbitrary sentence-length variation;\n- unusual synonyms chosen only to make wording less predictable;\n- metaphors that replace a precise technical explanation;\n- deliberate imperfections or tangents;\n- an invented AI probability or numerical slop score.\n\n## Preserve Markdown and product accuracy\n\n- Do not change fenced code, commands, URLs, link targets, image targets, or\n  frontmatter unless the request requires it.\n- Do not rename a heading if another page links to its generated anchor without\n  updating that link.\n- Do not change a UI label to improve prose. Rewrite the surrounding sentence.\n- Do not remove repetition that is required for independent reference sections.\n- Do not convert a walkthrough into reference documentation, or reference\n  documentation into a narrative tutorial, without user approval.\n- Do not infer product behaviour from the prose alone when the edit changes a\n  technical claim. Check the implementation or an authoritative reference.\n- Treat contradictions between prose, screenshots, capture metadata, sample\n  repositories, and implementation as accuracy defects. Resolve them from the\n  authoritative source instead of rewriting around them.\n- When a page title changes, update the sidebar, index, related-page labels, and\n  in-scope links that display the old title.\n\n## Preserve images and other media\n\nTreat every existing image, diagram, video, and asset as preserved content.\nHumanizing prose does not authorize media removal or replacement.\n\n- Do not delete an asset, remove its reference, change its order, or replace it\n  unless the user explicitly approves that action.\n- Do not use \"task value,\" brevity, a stale appearance, or a text explanation\n  as automatic reasons to remove an image.\n- Do not bulk-delete assets during a prose revision.\n- If an image is stale, private, inaccurate, decorative, or duplicated, report\n  the issue and propose one action: keep, recapture, move, or remove. Wait for\n  approval before changing it.\n- If an image exposes a credential or other active secret, stop publication and\n  report it immediately. Do not silently make a wider set of image changes.\n- When a replacement is approved, capture or obtain the replacement before\n  removing the current asset. Preserve the route and layout while replacement\n  work is pending.\n- Check non-doc consumers before changing an asset. README files, homepages,\n  package pages, and other applications can import docs-site images directly.\n\nA decision not to add a new screenshot is not permission to remove an existing\nscreenshot.\n\n## Control the size of the rewrite\n\nFor a large documentation set, work in reviewed vertical slices. Complete and\nreview one representative page before applying the approach to the rest of a\nslice. Do not perform a site-wide structural rewrite from an audit summary.\n\nPause and ask for approval when the work would:\n\n- change the dominant format of a page, such as prose to lists;\n- remove substantial explanation, examples, or media;\n- merge, tombstone, redirect, or delete a page;\n- change many pages through the same structural template; or\n- produce a much larger diff than the factual and prose defects require.\n\nWhen several agents contribute, give them the same preservation set and require\na central review of format balance and media changes before integration.\n\n## Use a two-pass edit\n\n### Pass 1: structure\n\nRemove redundant framing, merge repeated explanations, order information by the\nreader's task, and keep prerequisites before dependent actions. Give prominent\nplacement to features that materially improve repeated use; do not give every\nfeature equal weight merely because the source page did.\n\nKeep the smallest effective structural change. Do not rewrite a complete page\nwhen a heading, transition, or paragraph edit fixes the defect.\n\n### Pass 2: sentences\n\nRemove filler and AI mannerisms. Simplify grammar. Keep terminology and facts\nstable. Read the result as technical documentation, not as marketing copy.\n\nThen compare the result with the preservation set. Correct any lost condition,\nchanged claim, broken reference, altered example, or mismatched UI label.\nCompare the prose-to-list balance and media inventory with the original. Undo\nunrequested format conversions and media changes.\n\n## Validate edited files\n\nRun checks that match the change:\n\n1. Inspect the diff for accidental changes to code blocks, links, images,\n   examples, numbers, and UI labels.\n2. List every added, removed, moved, or replaced media asset. Confirm that each\n   change has explicit approval and that no external consumer breaks.\n3. Search for renamed heading anchors and update in-scope references when the\n   rename was intentional.\n4. If the tutorial uses a sample repository, verify its URL, files, commands,\n   expected failures, and required authentication.\n5. Run `git diff --check`.\n6. Run documentation build or link checks when the change can affect rendering\n   or navigation and the repository provides a focused command.\n7. When changes affect structure, rendering, navigation, or media, review at\n   least one rendered page from each changed slice. Check that prose still\n   explains the subject and that lists support rather than replace it.\n8. Inspect the diff summary. Stop and review again if it shows unexpected mass\n   deletions, repeated page rewrites, or asset removal.\n9. State exactly what was checked and what was not checked.\n\nReport the main structural changes. Do not paste the full rewritten document\nunless the user asks for it.\n\n## Example\n\nBefore:\n\n> Three things in that description are worth noticing, because each one becomes\n> a part of the plan you will see in a moment.\n\nAfter:\n\n> The description controls three parts of the plan:\n\nThe revision keeps the useful relationship and removes the tutorial\nannouncement. If the count is not useful, remove it too and lead with the first\ninstruction.\n","# Sero Humanize\n\nMake Sero prose direct, specific, and useful. Human writing in technical\ndocumentation does not need personality. It needs clear decisions, concrete\nfacts, and respect for the reader's time.\n\nApply this skill to Sero content even when drafting it in chat. Use `unslop`\nfor the surrounding conversation, not for the content itself.\n\n## Follow the requested mode\n\n- For an audit or review, report the material patterns and do not edit files.\n- For an edit, rewrite the named files in place.\n- For a new document, apply these rules while drafting it.\n- If the request does not specify a mode, infer it from the requested action.\n  Do not turn a request to assess prose into permission to change it.\n\n## Establish the voice and preservation set\n\nBefore editing:\n\n1. For a passage edit, read the passage and enough surrounding text to preserve\n   meaning. Read the full file for a whole-document audit or structural rewrite.\n2. Read the nearest repository instructions that apply to those files.\n3. Use adjacent, clearly human-edited Sero documentation as the voice sample\n   when the named files do not establish a consistent voice.\n4. Record what must not change:\n   - technical meaning and product behaviour;\n   - product names, canonical terms, and exact UI labels;\n   - commands, code, file paths, numbers, limits, and factual claims;\n   - the documentation type, useful narrative flow, and balance between prose,\n     lists, tables, examples, and callouts;\n   - frontmatter, anchors, link targets, image paths, and screenshot order;\n   - the purpose and placement of each image, diagram, and other media asset;\n   - quotations and user input examples, unless the user asks to edit them.\n\nIdentify the intended reader. Unless the page states otherwise, assume the\nreader knows neither Sero nor the feature. Do not assume that simpler grammar\nfixes an explanation that requires missing product knowledge.\n\nDo not add a fact to make a sentence more vivid. Verify a doubtful claim from\nthe repository or leave it unchanged and report the doubt.\n\n## Audit structure before wording\n\nLook for clusters and repeated patterns. Do not treat one punctuation mark or\none common word as proof of AI writing.\n\nPrioritize these defects:\n\n- Meta narration that announces the next explanation instead of giving it.\n- Repeated tutorial staging such as \"what you are about to learn\" and \"what you\n  have learned.\"\n- Fixed enumerations such as \"three things are worth noticing\" when a direct\n  heading or short list is clearer.\n- Several paragraphs that can change order without changing the argument.\n- A heading followed by a sentence that only repeats the heading.\n- A conclusion that repeats the introduction without adding an action or fact.\n- Repeated summaries of the same screen, process, or result.\n- Forced contrasts such as \"not only X, but Y\" or \"not X; rather Y.\"\n- Groups of three used for rhythm instead of meaning.\n- Mechanical bold lead-ins, excessive inline bold, or lists that should be\n  short prose.\n- Promotional adjectives, vague importance claims, and unsupported praise.\n- Vague actors, passive constructions, filler, stacked hedges, and abstract\n  nouns where an action is available.\n- Synonym cycling for one product concept. Repeat the canonical term.\n- Long sentences that mix instructions, exceptions, and background.\n- Em dashes used repeatedly to join thoughts that need separate sentences.\n- Headings that narrate the demo or expose implementation language instead of\n  naming the reader's task, such as \"The finish\" or \"Answer the gate.\"\n- Examples that depend on an unexplained demo domain and therefore do not help\n  the reader understand the feature.\n- Tutorials that start using the product before they give prerequisites,\n  sample data, expected starting state, or required sign-ins.\n- Result sections that recite one captured run instead of telling the reader\n  what to inspect and verify.\n\nKeep useful structure. A list, summary, warning, question heading, or em dash is\nnot a defect by itself.\n\nDo not normalize a page or site to one format. Humanize the defective passages,\nnot every paragraph. If a page already explains a concept well in prose, keep\nit as prose.\n\n## Rewrite for Sero documentation\n\nApply ASD-STE100 Simplified Technical English where it fits the material:\n\n- Put the action or answer first.\n- Use active voice when the actor matters.\n- Give one main instruction per sentence.\n- Put a condition before the action when the reader must know it first.\n- Prefer common, precise words over formal or promotional alternatives.\n- Use the same term for the same thing.\n- Keep paragraphs focused on one subject.\n- Keep necessary limits, cautions, and exceptions close to the action.\n- Use contractions only when the established local voice requires them.\n- Keep exact UI text in bold when the documentation uses bold for controls.\n- Keep code identifiers and paths in code formatting.\n- Retain a summary only when it helps the reader decide or act.\n\nUse lists only when the content is naturally a sequence, set of choices,\nchecklist, or compact reference. Do not:\n\n- convert explanatory prose into bullet points only to make it shorter;\n- turn each sentence or paragraph into a list item;\n- replace transitions and reasoning with disconnected bullets;\n- use repeated lists where a short paragraph gives the reader necessary\n  context; or\n- make several pages share the same mechanical list structure.\n\nAfter the sentence pass, read the page as a whole. If lists now dominate a page\nthat previously used useful prose, restore the prose. Clear technical writing\nneeds connected explanation as well as scannable reference material.\n\nFor an overview page:\n\n- Explain the feature in familiar words before using its product terms.\n- State what the user gives Sero, what Sero does, and what the user reviews.\n- When comparing features, give one plain decision rule. Use examples that a\n  reader can understand without knowing the tutorial repository or a specialist\n  software domain.\n\nFor a tutorial:\n\n- Put setup before the first product action. Include required software,\n  accounts, sign-ins, repository or sample-data setup, and a command or visible\n  result that confirms the expected starting state.\n- Prefer a stable sample repository over instructions that ask an agent to\n  generate approximate sample data. Verify the repository contents and commands\n  before documenting them.\n- Use task-based headings such as \"Review the plan,\" \"Change the plan,\" and\n  \"Check the result.\" A heading must describe the full purpose of its section;\n  do not narrow a general control to one example case.\n- End with checks the reader can perform. Do not use a captured run's cost,\n  duration, names, or outcome as a substitute for verification instructions.\n\nFor feature language:\n\n- Use the visible object name: icon, button, tab, question, or approval request.\n  Do not call an icon a mark or expose internal terms such as gate, fan-out, or\n  feedback route when plain behaviour is enough.\n- Keep exact UI labels unchanged, but explain them with common words.\n- Put high-value quality-of-life features where readers will find them. Give\n  them enough space to explain when the control appears, how to use it, what it\n  changes, and what remains under user control.\n\nCompress or merge only the passages that contain a verified structural defect.\nKeep useful depth, examples, transitions, and paragraph structure. A shorter\npage is not automatically a better page.\n\nDo not manufacture a human voice with:\n\n- anecdotes, opinions, jokes, sensory details, or personal asides;\n- fragments, one-word sentences, or dramatic punch lines;\n- arbitrary sentence-length variation;\n- unusual synonyms chosen only to make wording less predictable;\n- metaphors that replace a precise technical explanation;\n- deliberate imperfections or tangents;\n- an invented AI probability or numerical slop score.\n\n## Preserve Markdown and product accuracy\n\n- Do not change fenced code, commands, URLs, link targets, image targets, or\n  frontmatter unless the request requires it.\n- Do not rename a heading if another page links to its generated anchor without\n  updating that link.\n- Do not change a UI label to improve prose. Rewrite the surrounding sentence.\n- Do not remove repetition that is required for independent reference sections.\n- Do not convert a walkthrough into reference documentation, or reference\n  documentation into a narrative tutorial, without user approval.\n- Do not infer product behaviour from the prose alone when the edit changes a\n  technical claim. Check the implementation or an authoritative reference.\n- Treat contradictions between prose, screenshots, capture metadata, sample\n  repositories, and implementation as accuracy defects. Resolve them from the\n  authoritative source instead of rewriting around them.\n- When a page title changes, update the sidebar, index, related-page labels, and\n  in-scope links that display the old title.\n\n## Preserve images and other media\n\nTreat every existing image, diagram, video, and asset as preserved content.\nHumanizing prose does not authorize media removal or replacement.\n\n- Do not delete an asset, remove its reference, change its order, or replace it\n  unless the user explicitly approves that action.\n- Do not use \"task value,\" brevity, a stale appearance, or a text explanation\n  as automatic reasons to remove an image.\n- Do not bulk-delete assets during a prose revision.\n- If an image is stale, private, inaccurate, decorative, or duplicated, report\n  the issue and propose one action: keep, recapture, move, or remove. Wait for\n  approval before changing it.\n- If an image exposes a credential or other active secret, stop publication and\n  report it immediately. Do not silently make a wider set of image changes.\n- When a replacement is approved, capture or obtain the replacement before\n  removing the current asset. Preserve the route and layout while replacement\n  work is pending.\n- Check non-doc consumers before changing an asset. README files, homepages,\n  package pages, and other applications can import docs-site images directly.\n\nA decision not to add a new screenshot is not permission to remove an existing\nscreenshot.\n\n## Control the size of the rewrite\n\nFor a large documentation set, work in reviewed vertical slices. Complete and\nreview one representative page before applying the approach to the rest of a\nslice. Do not perform a site-wide structural rewrite from an audit summary.\n\nPause and ask for approval when the work would:\n\n- change the dominant format of a page, such as prose to lists;\n- remove substantial explanation, examples, or media;\n- merge, tombstone, redirect, or delete a page;\n- change many pages through the same structural template; or\n- produce a much larger diff than the factual and prose defects require.\n\nWhen several agents contribute, give them the same preservation set and require\na central review of format balance and media changes before integration.\n\n## Use a two-pass edit\n\n### Pass 1: structure\n\nRemove redundant framing, merge repeated explanations, order information by the\nreader's task, and keep prerequisites before dependent actions. Give prominent\nplacement to features that materially improve repeated use; do not give every\nfeature equal weight merely because the source page did.\n\nKeep the smallest effective structural change. Do not rewrite a complete page\nwhen a heading, transition, or paragraph edit fixes the defect.\n\n### Pass 2: sentences\n\nRemove filler and AI mannerisms. Simplify grammar. Keep terminology and facts\nstable. Read the result as technical documentation, not as marketing copy.\n\nThen compare the result with the preservation set. Correct any lost condition,\nchanged claim, broken reference, altered example, or mismatched UI label.\nCompare the prose-to-list balance and media inventory with the original. Undo\nunrequested format conversions and media changes.\n\n## Validate edited files\n\nRun checks that match the change:\n\n1. Inspect the diff for accidental changes to code blocks, links, images,\n   examples, numbers, and UI labels.\n2. List every added, removed, moved, or replaced media asset. Confirm that each\n   change has explicit approval and that no external consumer breaks.\n3. Search for renamed heading anchors and update in-scope references when the\n   rename was intentional.\n4. If the tutorial uses a sample repository, verify its URL, files, commands,\n   expected failures, and required authentication.\n5. Run `git diff --check`.\n6. Run documentation build or link checks when the change can affect rendering\n   or navigation and the repository provides a focused command.\n7. When changes affect structure, rendering, navigation, or media, review at\n   least one rendered page from each changed slice. Check that prose still\n   explains the subject and that lists support rather than replace it.\n8. Inspect the diff summary. Stop and review again if it shows unexpected mass\n   deletions, repeated page rewrites, or asset removal.\n9. State exactly what was checked and what was not checked.\n\nReport the main structural changes. Do not paste the full rewritten document\nunless the user asks for it.\n\n## Example\n\nBefore:\n\n> Three things in that description are worth noticing, because each one becomes\n> a part of the plan you will see in a moment.\n\nAfter:\n\n> The description controls three parts of the plan:\n\nThe revision keeps the useful relationship and removes the tutorial\nannouncement. If the count is not useful, remove it too and lead with the first\ninstruction.\n",{"name":5,"description":7},"main","e91c3c773b84e88a929cae3c36977e892ca4871a",".agents\u002Fskills\u002Fsero-humanize",4,"https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero","https:\u002F\u002Fsero-ai.dev\u002F","https:\u002F\u002Fgithub.com\u002Fsero-labs","2026-11-10 06:55:21",{"source":8,"skillName":5,"commands":63,"prompt":73,"zipAvailable":74},[64,67,70],{"tool":65,"command":66},"npx","npx skills add sero-labs\u002Fsero --skill sero-humanize",{"tool":68,"command":69},"bunx","bunx skills add sero-labs\u002Fsero --skill sero-humanize",{"tool":71,"command":72},"pnpm","pnpm dlx skills add sero-labs\u002Fsero --skill sero-humanize","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fsero-humanize\u002FSKILL.md",true,[76,82,86,90,94,98,102,106,110,114,118,122,126,130],{"path":77,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":80,"sourceHtmlUrl":81},"AGENTS.md","REFERENCE","参考文档",22627,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002FAGENTS.md",{"path":83,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":84,"sourceHtmlUrl":85},"agents\u002Fopenai.yaml",212,"https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fsero-humanize\u002Fagents\u002Fopenai.yaml",{"path":87,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":88,"sourceHtmlUrl":89},"metadata.json",530,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Fmetadata.json",{"path":91,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":92,"sourceHtmlUrl":93},"README.md",2140,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002FREADME.md",{"path":95,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":96,"sourceHtmlUrl":97},"rules\u002F_sections.md",829,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002F_sections.md",{"path":99,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":100,"sourceHtmlUrl":101},"rules\u002F_template.md",329,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002F_template.md",{"path":103,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":104,"sourceHtmlUrl":105},"rules\u002Farchitecture-avoid-boolean-props.md",2267,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Farchitecture-avoid-boolean-props.md",{"path":107,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":108,"sourceHtmlUrl":109},"rules\u002Farchitecture-compound-components.md",2600,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Farchitecture-compound-components.md",{"path":111,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":112,"sourceHtmlUrl":113},"rules\u002Fpatterns-children-over-render-props.md",1886,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Fpatterns-children-over-render-props.md",{"path":115,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":116,"sourceHtmlUrl":117},"rules\u002Fpatterns-explicit-variants.md",2395,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Fpatterns-explicit-variants.md",{"path":119,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":120,"sourceHtmlUrl":121},"rules\u002Freact19-no-forwardref.md",953,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Freact19-no-forwardref.md",{"path":123,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":124,"sourceHtmlUrl":125},"rules\u002Fstate-context-interface.md",4974,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Fstate-context-interface.md",{"path":127,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":128,"sourceHtmlUrl":129},"rules\u002Fstate-decouple-implementation.md",2699,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Fstate-decouple-implementation.md",{"path":131,"resourceType":78,"resourceTypeDesc":79,"sizeBytes":132,"sourceHtmlUrl":133},"rules\u002Fstate-lift-state.md",3224,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-composition-patterns\u002Frules\u002Fstate-lift-state.md",[135,167,194,232,259,284],{"skillNo":136,"name":137,"nameZh":6,"description":138,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":139,"categories":140,"keywords":147,"publishedAt":49},"01M70PSGDVAGMAGE2RFKEQ1EW4","agent-browser","Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to \"open a website\", \"fill out a form\", \"click a button\", \"take a screenshot\", \"scrape data from a page\", \"test this web app\", \"login to a site\", \"automate browser actions\", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fagent-browser\u002FSKILL.md",[141,144],{"categoryNo":142,"categoryName":143,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":145,"categoryName":146,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000023","Skill与智能体开发",[148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,137,165,166],"agent browser","browser","testing","browser automation","notion","qa","figma","actions","messages","open","web tools","app","click","bedrock","extracting","conversations","programmatic","slack messages","exploratory testing",{"skillNo":168,"name":169,"nameZh":6,"description":170,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":171,"sourceHtmlUrl":172,"categories":173,"keywords":179,"publishedAt":49},"01M6CHW2A115ZZDEFNXV2Z7A2Z","skill-creator","Guide for writing and improving Sero skills. Use when the user wants to create a new skill, rewrite an existing one, or turn a repeated piece of work into reusable procedural knowledge for the agent.","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Fskill-creator\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Fskill-creator\u002FSKILL.md",[174,177,178],{"categoryNo":175,"categoryName":176,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":145,"categoryName":146,"taxonomyType":20,"taxonomyTypeDesc":21},{"categoryNo":23,"categoryName":24,"taxonomyType":25,"taxonomyTypeDesc":26},[180,181,48,182,183,184,185,186,34,187,188,189,190,31,191,192,193],"skill creator","creator","create","knowledge","work","user","guide","reusable","new skill","rewrite","procedural","Existing","when","new",{"skillNo":195,"name":196,"nameZh":6,"description":197,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":198,"sourceHtmlUrl":199,"categories":200,"keywords":210,"publishedAt":49},"01M6CHVJCZS9V3WWFPG47WC4G9","frontend-design","Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML\u002FCSS layouts, or when styling\u002Fbeautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics.","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Ffrontend-design\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Ffrontend-design\u002FSKILL.md",[201,204,207],{"categoryNo":202,"categoryName":203,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000009","前端开发",{"categoryNo":205,"categoryName":206,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":208,"categoryName":209,"taxonomyType":25,"taxonomyTypeDesc":26},"01JBSEEDCAT000000000000032","设计师",[211,212,213,214,215,216,217,218,219,182,42,220,30,221,222,223,224,225,34,226,227,33,228,229,230,231,192],"frontend","design","frontend design","css","web","components","react","react components","web components","artifacts","examples","applications","dashboards","quality","landing","html","styling","ui design","build","aesthetics","generic",{"skillNo":233,"name":234,"nameZh":6,"description":235,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":236,"categories":237,"keywords":239,"publishedAt":49},"01M6CHVXWDN98CP3MPT2WZ1ABN","sero-plugin","Create Sero plugins — self-contained packages with a Pi extension (agent\ntools\u002Fcommands), an optional React UI loaded via Module Federation, and an\noptional background runtime. Use when the user asks to create a Sero\nplugin\u002Fapp\u002Fextension, add a sidebar panel, dashboard widget, or\nagent-integrated feature, or convert a Pi extension into a Sero plugin.\nTrigger on phrases like \"create a plugin\", \"new Sero app\", \"add a tool\nwith UI\", \"build an extension\", \"convert this Pi extension\", or any\nrequest that touches `plugins\u002Fsero-*-plugin\u002F`.","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Fsero-plugin\u002FSKILL.md",[238],{"categoryNo":145,"categoryName":146,"taxonomyType":20,"taxonomyTypeDesc":21},[240,241,242,217,243,182,244,30,245,246,247,248,249,185,250,34,229,251,252,253,254,255,159,256,257,258,192,193],"plugin","sero plugin","dashboard","packages","plugins","federation","Tools","react ui","tool","convert","pi","self","commands","runtime","pi extension","feature","add","panel","request",{"skillNo":260,"name":261,"nameZh":6,"description":262,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":263,"categories":264,"keywords":272,"publishedAt":49},"01M6CHVTJ5Y9JE6181SVPQXNZY","sero-dashboard-ui","Build, redesign or standardise Sero dashboard widgets and compact plugin\nviews using the shared @sero-ai\u002Fui dashboard components. Use when the user\nasks to create, restyle, clean up or make consistent a dashboard widget or a\ndata-dense plugin view. Trigger on phrases like \"dashboard widget\", \"widget\nfor the dashboard\", \"restyle this widget\", \"make the widget consistent\",\n\"compact plugin view\", or any work under a plugin's `ui\u002Fwidgets\u002F`.","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Fsero-dashboard-ui\u002FSKILL.md",[265,268,269],{"categoryNo":266,"categoryName":267,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000013","分析与可视化",{"categoryNo":208,"categoryName":209,"taxonomyType":25,"taxonomyTypeDesc":26},{"categoryNo":270,"categoryName":271,"taxonomyType":25,"taxonomyTypeDesc":26},"01JBSEEDCAT000000000000033","数据分析师",[242,30,273,216,240,182,274,184,275,276,185,34,277,278,229,279,280,281,282,283,31,192],"sero dashboard ui","data","redesign","view","up","clean","compact","shared","widgets","trigger","make",{"skillNo":285,"name":286,"nameZh":6,"description":287,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":288,"categories":289,"keywords":291,"publishedAt":49},"01M6CHVR09BNM75VJXMQN0G7MG","sero-browser","Use when a user asks to open, browse, capture, screenshot, or record a website\u002Fpage; mentions browser tabs, visible Browser panel, screen recording, web navigation, website interaction, or browser automation; or when choosing between `sero browser`, `sero app record`, and `automation_browser`.","https:\u002F\u002Fgithub.com\u002Fsero-labs\u002Fsero\u002Fblob\u002Fmain\u002Fpackages\u002Ftemplates\u002Fskills\u002Fsero-browser\u002FSKILL.md",[290],{"categoryNo":145,"categoryName":146,"taxonomyType":20,"taxonomyTypeDesc":21},[149,292,151,293,215,294,295,296,297,157,298,299,185,34,300,301,159,302,303,257,304,305,192],"sero browser","automation","screenshot","website","screen recording","browse","page","capture","navigation","record","interaction","screen","web navigation","recording"]