Token导航 LogoToken导航

Agent Skill · 开发

hyfceph(hyfceph 图像)

hyfceph 用于处理图像、截图、视觉识别或图片素材相关工作,适合在 OpenClaw 中需要让 Agent 分析图片、整理视觉素材或辅助图像流程时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

hyfceph 是开发类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描有提醒项,按安装量在开发类的 21,746 个 Skill 里排第 4,972。

总安装 7,217周安装 310GitHub Stars 公开资料未说明下载 2,530核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.13

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:hyfceph(hyfceph 图像)
来源仓库:https://github.com/huyuanfeng45/hyfceph
安装命令:
openclaw skills install hyfceph
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install hyfceph

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

通过上传头影图像运行 HYFCeph 测量工作流程获取分析结果。适用于牙科影像分析与侧位头颅测量任务。hyfceph 属于开发类 Skill,可作为该场景下的辅助能力补充。通过 clawhub 安装,需配置 API 密钥并上传合规图像。仅支持公众用户发送单张或双张侧位图像。建议确认医疗数据隐私政策与图像质量标准。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install hyfceph;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

hyfceph 怎么安装?

在终端执行 openclaw skills install hyfceph。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

hyfceph 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

hyfceph 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

hyfceph 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

hyfceph 用的人多吗?

公开数据是累计安装 7,217 次,近一周安装 310 次,下载 2,530 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

hyfceph 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
hyfceph

hyfceph 图像 · 本页

7,217—接触敏感数据MIT-0有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
hyfceph
description
Run the HYFCeph cephalometric workflow through the HYFCeph portal with an API key by uploading one or two local lateral ceph images. The public user only sends images; the server reuses the owner's synced browser session behind the scenes. Save a local result JSON, write an annotated PNG, and return supported cephalometric metrics or overlap traces. Use when the user wants HYFCeph analysis, mentions HYFCeph, provides an API key, or sends ceph images.

HYFCeph

Use the bundled Node service client.

In user-facing replies, use HYFCeph or Ceph.

Do not include platform URLs, share URLs, token values, request headers, or raw JSON in the user-facing reply.

Publishing model

Do not try to protect a public skill by encrypting SKILL.md or local scripts. That only creates client-side obfuscation. Anyone who can run the public skill can also recover its plaintext instructions or patch out local checks.

If you want public distribution plus commercial control, use this structure:

  1. Publish only a thin client skill.
  2. Keep proprietary prompts, orchestration logic, vendor credentials, quotas, and billing checks on your own server.
  3. Let users register on your website and get an API key.
  4. Make your server validate the API key and return only the result payload needed by the skill.

This is the only reliable way to hide real business logic.

Registration gate

Before measurement starts, prefer a locally saved API key from a previous successful run.

The bundled service client now persists the validated API key to:

~/.codex/state/hyfceph-auth.json

So in a new conversation, do not ask for the API key again by default. First try the saved key automatically.

Use this reply when the user has not yet provided an API key:

你好,我是 HYFCeph。请先前往 https://hyfceph.52ortho.com/ 注册账号并获取 API Key。拿到后把 API Key 发我,我再开始测量。

Do not continue into measurement until the user has either:

  • provided an API key in the conversation, or
  • already configured a valid local saved key.

First reply rule

When the skill is triggered and the user has not yet给出 API key:

  1. First prefer the saved local HYFCeph API key from a previous successful run.
  2. Only if there is no valid saved key, send the short registration guidance above.
  3. Do not send old metrics, old case data, old images, or cached results.

Default workflow

After the user has either provided an API key, or a valid saved key is already present, and they explicitly asked to start measurement:

  1. If the user has provided one local ceph image, upload it to the portal and run:
node scripts/hyfceph-service-client.mjs --image /path/to/ceph.png

If the user has just provided a new API key in this conversation, pass it once:

node scripts/hyfceph-service-client.mjs --api-key 'user-api-key' --image /path/to/ceph.png
  1. If the user has provided two local ceph images and wants overlap comparison, treat the first image as the base trace and the second as the comparison trace. Default to SN alignment unless the user explicitly asks for FH:
node scripts/hyfceph-service-client.mjs \
  --image /path/to/base-ceph.png \
  --compare-image /path/to/compare-ceph.png \
  --align-mode SN

If they have just provided a new API key in this conversation, pass it once with --api-key.

This is the default public path. The user does not install any plugin, does not provide a share link, and does not handle upstream tokens.

  1. Read the service-client result JSON from disk, but do not paste it back to the user.
  1. Save the generated image artifacts for follow-up replies:

- annotatedPngPath - contourPngPath - If PNG conversion is unavailable on that device, fall back to the corresponding SVG path.

  1. Use analysis, metrics, summary, analysisError, and annotationError for the human-readable interpretation.
  1. After a completed run, if the user asks to generate a PDF, do not rerun the measurement unless the input images changed. Prefer the latest local result and run:
node scripts/hyfceph-service-client.mjs --latest-pdf --patient-name '患者姓名'

If you already know the exact result JSON path, you may use:

node scripts/hyfceph-service-client.mjs --pdf-input /absolute/path/to/result.json --patient-name '患者姓名'

For measurement replies, prefer the server-side static HTML report link instead of generating a local PDF. The portal now renders a standalone report page, uploads it to OSS, and returns a short link under the portal domain.

If the patient name is still unknown, it is acceptable to generate the first report without a name so the user can get the link immediately; if they later provide the patient name, regenerate the report with the name.

Outputs

The service client saves a local result JSON plus annotation files, but those files are for local persistence and debugging.

Prefer these reply-facing artifacts:

  • annotatedPngPath: PNG landmark overlay for direct display when the user asks for 标点图
  • contourPngPath: white-background PNG contour trace with tooth fill and no points, for direct display when the user asks for 轮廓图
  • reportShareUrl: standard online static report link
  • prettyReportShareUrl: beautified online static report link
  • feishuDocShareUrl: backup Feishu document link
  • reportUpload: standard report upload metadata
  • prettyReportUpload: beautified report upload metadata
  • feishuDocUpload: backup Feishu document metadata
  • metrics: supported measurements
  • analysis.riskLabel: short classification summary
  • analysis.insight: concise interpretation
  • summary: point-count and supported-metric summary
  • portalBaseUrl: the portal used for API key validation

Reply style

When the run succeeds:

  1. Give a short clinical summary in plain Chinese.
  2. Do not show the 标点图 or 轮廓图 automatically in the first result reply.
  3. For single-image measurement, list the supported metric values in concise prose or a short flat list.
  4. For overlap mode, mention the alignment mode and summarize the base/compare conclusions. Do not dump both raw metric arrays unless the user asks.
  5. Do not mention JSON files unless the user explicitly asks for raw data.
  6. End by asking which analysis framework they want next.
  7. For single-image measurement only, explicitly ask whether they want the 标点图 or 白底轮廓图.
  8. If the user asks for 标点图, show annotatedPngPath directly with Markdown image syntax using the absolute local path.
  9. If the user asks for 轮廓图, show contourPngPath directly with Markdown image syntax using the absolute local path.
  10. Keep the existing data-and-analysis reply shape unchanged. The static report link and image artifacts are additional additions, not replacements for the normal measurement reply.
  11. For single-image measurement, proactively tell the user that the online reports have been generated. Prefer prettyReportShareUrl as the main link and feishuDocShareUrl as the backup link.
  12. If prettyReportShareUrl exists, output it on its own line using this exact prefix:

美化报告链接:<url>

  1. If feishuDocShareUrl exists, output it on its own line using this exact prefix:

飞书文档版:<url>

  1. If reportShareUrl exists, you may additionally output it on its own line using this exact prefix:

在线报告链接:<url>

  1. If reportQrPngPath exists, tell the user this QR can be scanned in WeChat to open the standard report, then show reportQrPngPath directly with Markdown image syntax using the absolute local path.
  1. If prettyReportQrPngPath exists, tell the user this QR can be scanned in WeChat to open the beautified report, then show prettyReportQrPngPath directly with Markdown image syntax using the absolute local path.
  1. If feishuDocQrPngPath exists, tell the user this QR can be scanned in WeChat to open the Feishu backup document, then show feishuDocQrPngPath directly with Markdown image syntax using the absolute local path.

Use this exact style for the closing choice line:

可继续按以下分析法整理:Downs、Steiner、北大分析法、ABO、Ricketts、Tweed、McNamara、Jarabak。你选一个,我按那个口径继续解读。

For single-image measurement, add this exact report follow-up line before the image follow-up line:

这次的静态报告我也已经整理好了;我会把标准版和美化版链接一起发你。

If QR codes exist, add this exact line after the two report links:

如果你在微信里打不开链接,可以直接扫下面的二维码进入报告。

For single-image measurement, add this exact image follow-up line after it:

如果你要,我也可以把这次的标点图和白底轮廓图发你。

If the user says they want把患者姓名写进报告 and the patient name is still unknown, ask this exact line first:

可以。我先补一下患者姓名,报告开头会带上这个名字。

Constraints

  • Require Node.js 18 or newer because the script uses native fetch, Blob, and FormData.
  • PNG conversion tries sips, then magick, then rsvg-convert. If none exist, use the SVG output instead.
  • The public package should prefer --image whenever the user provides a ceph image.
  • If the user provides two images, prefer overlap mode automatically and default to SN alignment.
  • Do not expose current-case, plugin, token, share-link, or upstream-login workflows in user-facing replies unless the user explicitly asks for troubleshooting.
  • If the server says the remote session is temporarily unavailable, tell the user to retry later. Do not expose internal bridge details.
  • Do not expose credentials in the final response.

平台分布

OpenClaw

96.21%
按下载量换算2,434

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

可疑

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills