**个人健康档案台(离线隐私版)**

SKILL开发
开发 未标注权限 许可证不明确
安装这个技能

离线、零依赖、数据全在本地浏览器的个人健康档案台:指标趋势图 + 文字类结果按检查部位分组比对,全部记录可编辑。

一句话选型结论

按资料字段整理

**个人健康档案台(离线隐私版)**是开发类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类开发 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_45371586/single-file-html-editable
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

开发
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_45371586仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791625067
正文体量12.8 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
写代码与改 bug

按需求补代码、定位并修掉报错。

读仓库与生成改动

让模型读懂项目结构,再给出可落地的改动。

跑测试与排查故障

按现象排查问题,补测试用例并复跑。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在开发场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
**个人健康档案台(离线隐私版)**当前 未标注权限 未标注 未标注 未标注 有
技能审查员 浏览器控制 ClawHub 市场 877.9 万 OpenClaw 有
Vercel React 最佳实践 浏览器控制 GitHub 仓库 849.1 万 Codex、Claude、Cursor 有
GitHub 代码协作 联网访问 ClawHub 市场 763.3 万 OpenClaw 有
网页设计指南 联网访问 GitHub 仓库 668.6 万 Codex、Claude、Cursor 有
本体论 只读 ClawHub 市场 663.7 万 OpenClaw 有

SKILL.md 全文

12.8 KB
模型实际读到的说明书原文,来自技能仓库

Single-File HTML Editable Pattern

Overview

This skill captures a repeatable pattern for taking a single-file, offline, localStorage-backed
HTML app (vanilla JS, no build, no server, no CDN) and adding two capabilities that are tedious
to get right by hand:

  1. Record-level editing — every stored record gets an "编辑" button that backfills the input
    form, switches the primary button to "保存修改", offers a "取消", and saves in place by id
    instead of creating a duplicate.
  2. Grouped longitudinal comparison for text-only types — for record types that are free-text
    (no numeric sub-fields, hence no trend chart), group records by a chosen field (e.g. 检查部位 /
    exam site) with a filter dropdown, so repeated exams of the same site line up chronologically.

It also mandates a strict jsdom verification step (real click/change events, dates generated
dynamically so tests are stable on any run date, simulated-reload persistence checks) and a
byte-identical skill-sync discipline so a shipped skill copy is provably the same file that was
verified. Use it when building a new such app or augmenting an existing one.

When to use

  • A request to make entries in a standalone HTML tracker editable ("让记录能改", "加编辑功能", "记录可编辑").
  • A request to compare free-text records over time ("文字结果比对不方便", "按部位看变化", "纵向比对").
  • A request about a 个人健康档案 / 健康档案 / 健康记录 style app — either generate a new one from
    the bundled template (assets/health-archive/index.html, verified) or augment an existing
    single-file app with editing/text grouping. e.g. "做一个个人健康档案台", "给个人健康档案加编辑功能",
    "健康档案里的 B超结果按部位比对".
  • Distributing a single-file HTML template as a WorkBuddy skill and keeping its bundled copy
    verifiably identical to the tested work copy.

Bundled example: 个人健康档案台

assets/health-archive/ ships the full worked example that this pattern was distilled from:
a single-file, offline 个人健康档案台 (personal health archive dashboard) with 7 modules
(体检指标趋势 / 用药·就诊 / 每日体征打卡 / 报告附件归档 / 复查提醒 / 家庭病史 / 过敏史),
all records editable in place, and text-only imaging results (B超/核磁/CT) grouped by 检查部位
with a filter dropdown. Its verify.cjs runs 194 jsdom checks and is the reference
implementation of every step in this skill. Use it as a starting point or to study the pattern
applied end-to-end.

Do NOT use this for apps with a backend/API, a multi-file build system, or a framework that already
provides CRUD — the pattern targets the no-dependency, vanilla-JS, localStorage case.

Core pattern: edit state

Apply these steps per module that stores an array of records under a key in the app's DB
object (e.g. DB.meds, DB.metrics). The reference file
references/edit-state-pattern.md holds the full copy-pasteable code.

1. Global edit state + per-module cancel button

let editing = null;   // null = "add" mode; {id, kind} = currently editing

In each module's input card, add a hidden "取消" button next to the primary add button:

<div class="btn-col"><button class="primary" id="d-add">添加</button>
  <button id="d-cancel" style="display:none">取消</button></div>

2. Toggle add/edit UI appearance

Keep a map kind -> [addBtnId, cancelBtnId]. setEditMode(kind, on) flips the add button's
label to "保存修改" (caching the original in dataset.add) and shows/hides the cancel button.
cancelEdit() resets editing, flips all modules back to add mode, and clears the forms.

3. Add an edit button column to the list renderer

If using a shared renderRows(container, rows, cols, onDel, onEdit) helper, render an extra
<button data-edit="id">编辑</button> column when onEdit is supplied and call onEdit(r) on
click. For custom renderers (charts, cards), add the same data-edit button inside each row and
wire its click to startEditX(r.id).

4. Backfill the form on edit start

startEditX(id) finds the record, sets editing = {id, kind:"X"}, assigns each form field from
the record, then setEditMode("X", true). For modules with a type selector that rebuilds fields
(such as the metric picker), switch the selector to the record's type before backfilling the
type-specific fields.

5. Save handler updates in place

const e = (editing && editing.kind==="X") ? DB.X.find(r=>r.id===editing.id) : null;
if(e){ Object.assign(e, { /* new field values */ }); }   // id preserved
else  { DB.X.push({ id:uid(), /* new record */ }); }
editing = null; setEditMode("X", false); /* reset form */
save(); renderAll();

The invariant that separates "edit" from "add": when editing is set, the handler mutates the
existing record by id (array length unchanged) instead of pushing a new one.

Module-specific quirks (record these when they appear)

  • Date-unique records (daily check-ins): on edit save, if the new date collides with another
    record's date, reject with a toast and stay in edit mode (setEditMode(kind,true)); never
    overwrite the other record.
  • File-backed records (report attachments): the file is a base64 blob on the record. Editing
    changes metadata (name/date); provide a separate "重新选择文件" that replaces data/mime/size.
    FileReader.readAsDataURL is async — test it with a short await.
  • Records with history (reminders): editing must preserve done/history; only update visible
    fields, never reset completion state.

Core pattern: grouped text comparison

For a record type flagged cfg.textOnly (free text, no chart), replace the flat reverse-chron list
with grouping by a field:

if(cfg.textOnly){
  const groups = {};
  rows.forEach(r=>{ const site=(r.vals&&r.vals.site)||"未注明部位"; (groups[site]=groups[site]||[]).push(r); });
  const sites = Object.keys(groups).sort();
  let html = '<select id="textFilter"><option value="全部" selected>全部</option>' +
    sites.map(s=>'<option value="'+esc(s)+'">'+esc(s)+' ('+groups[s].length+')</option>').join("") + '</select>';
  const shown = textFilter==="全部" ? sites : [textFilter];
  shown.forEach(s=>{
    const g=groups[s], dates=g.map(r=>r.date).sort();
    const range = dates.length>1 ? dates[0]+" ~ "+dates[dates.length-1] : dates[0];
    html += '<div class="grp"><div class="grp-h">'+esc(s)+' <span class="tag">'+g.length+' 条</span>'+
            ' <span class="hint">'+esc(range)+'</span></div>' + recordCards(g, cfg) + '</div>';
  });
  el.innerHTML = html;
  const tf = document.getElementById("textFilter");
  if(tf) tf.addEventListener("change", ()=>{ textFilter=tf.value; renderMetricChart(); });
  return;
}

Rules:
- Keep a module-scoped let textFilter="全部"; and reset it to "全部" whenever the record-type
selector changes (so switching to a different type starts unfiltered).
- Group headers show the field value + count + date range — that is what makes "same site, over
time" readable at a glance.
- The card renderer (recordCards) reverses internally to newest-first; pass the group in natural
order and do not double-reverse.

Strict verification (jsdom)

Verification must be real, not asserted-by-faith. Use the bundled scripts/verify_harness.cjs as
the starting harness. It provides, out of the box:

  • A jsdom loader that executes the page's inline scripts (runScripts:"dangerously",
    pretendToBeVisual:true, url set so localStorage works).
  • Event helpers ev() (click) and chg() (change).
  • read() to parse the app's localStorage JSON (set the key constant).
  • Dynamic date helpers stable on any run date: T0, shiftDays(s,n), lastDayOfMonth(s), and
    addMonths(s,n) with month-end clamping (e.g. 08-31 + 1mo = 09-30, computed in Date.UTC
    to avoid timezone off-by-one).
  • clickEditIn(listId, id) to locate and click a record's edit button.
  • setFile(content, name, mime) to inject a File into <input type=file> and fire change for
    file-replacement testing (await a short timeout for the async FileReader).
  • byteIdentical(a, b) for the sync check.
  • A jsErrors collector and a universal "no JS error on load" assertion.

For every edit feature assert at minimum: (a) clicking 编辑 backfills the form and flips the button
to "保存修改"; (b) saving updates the existing record by id — array length unchanged, id
unchanged, the changed field updated; (c) clicking 取消 leaves data untouched; (d) for date-unique
types, a colliding date is rejected and edit mode persists; (e) for file types, replacing the file
updates data/mime/size; (f) after a simulated reload (re-load the DOM from saved
localStorage), the edit result and the grouping still render.

Skill-sync discipline (when shipping as a skill)

  1. Keep the work copy (e.g. health-archive/index.html + _verify.cjs) as the only place you
    edit and verify.
  2. After verify.cjs is green, cp both files into the skill directory and run diff -q to
    confirm byte-identical.
  3. Run verify.cjs inside the skill directory for end-to-end acceptance — this proves the shipped
    copy passes, not just the work copy.
  4. Update the skill's SKILL.md: module list, data-model buckets, behavior notes (edit + grouping),
    and the verification count.

Resources

  • scripts/verify_harness.cjs — reusable jsdom test harness with all helpers above; copy it next to
    the app's index.html, set the localStorage key and any app-specific element ids, then add
    per-app assertions in the marked block.
  • references/edit-state-pattern.md — full copy-pasteable code for the edit-state skeleton, the
    grouped-text renderer, the date utilities, and the exact module quirks, with the surrounding
    renderRows / renderMetricChart context.

Installation(安装 / 发给别人后怎么装)

这个 skill 本质上是一个目录,分发出去后放到本机的 skill 目录,WorkBuddy 就能直接唤起。
下面两种装法效果完全一致,选顺手的即可。

方式 A:从 zip 解压(推荐给接收方)

  1. 拿到 single-file-html-editable.zip。
  2. 解压到 WorkBuddy 的 skills 目录(不要多套一层文件夹,解压后应是 single-file-html-editable/SKILL.md):
    - Windows:把 zip 内容解压进 %USERPROFILE%\.workbuddy\skills\
    - macOS / Linux:解压进 ~/.workbuddy/skills/
  3. 重启 WorkBuddy(或刷新技能列表),用「个人健康档案」「给单文件 HTML 加编辑」「健康档案里的 B超按部位比对」等词即可唤起。

方式 B:直接拷目录

把 single-file-html-editable/ 整个文件夹(含 SKILL.md、references/、scripts/、assets/)复制到上面的 skills/ 目录即可,效果与方式 A 相同。

验证装上了

  • 唤起测试:用一句话试一下,例如「做一个个人健康档案台」「给单文件 HTML 加编辑功能」。
  • 或直接检查技能目录下是否出现 single-file-html-editable/SKILL.md。
  • 想跑自带验收:进入 skills/single-file-html-editable/assets/health-archive/,执行
    node verify.cjs(需先 npm i jsdom 到该目录或全局),看到 194/194 PASS 即模板完好。

常见问题

5 问
答案由本站依据条目字段整理
**个人健康档案台(离线隐私版)**怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_45371586/single-file-html-editable。安装方式为未标注。装完刷新技能列表或重开客户端。

**个人健康档案台(离线隐私版)**支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

**个人健康档案台(离线隐私版)**权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

**个人健康档案台(离线隐私版)**收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

**个人健康档案台(离线隐私版)**用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。