Contactless Health Risk Screening Tool | 非接触式健康风险检测分析工具

Combines frontal facial image capture with multimodal physiological feature analysis to provide early risk screening and alerts for chronic and acute conditi...

Security Scan

The skill's stated purpose (contactless facial health screening) matches its code, but there are several inconsistencies — undeclared environment/config access, local persistence and file writes, and automatic uploads of sensitive images to external APIs — that warrant caution before installing or using it.

Purpose & Capability

The name and description (face-based, non-contact health risk screening) align with the included code: face_analysis and smyx_common modules implement API calls, file validation, and result formatting. That capability legitimately requires reading images and calling external AI APIs. However, the skill includes a local DAO (SQLite) and config management utilities that create/read config files and persist data under the workspace; these are not called out in the manifest or SKILL.md as expected behaviors, creating a mismatch between declared requirements and actual capabilities.

Instruction Scope

SKILL.md explicitly forbids reading local memory and long-term memory, yet the code provides DAO/SQLite logic, a path-based get_db_path that writes under ${OPENCLAW_WORKSPACE}/data, and logic that saves uploaded attachments to the skill's attachments directory. The runtime instructions require saving user uploads and calling remote APIs with images; those network calls are coherent with the purpose but involve transmitting highly-sensitive biometric data to external endpoints. The SKILL.md enforces strict open-id retrieval rules (via config files) but the code also reads environment variables (OPENCLAW_SENDER_OPEN_ID, OPENCLAW_SENDER_USERNAME, FEISHU_OPEN_ID) and will create or load YAML config files — behavior not declared in the manifest.

Install Mechanism

No install spec is provided (instruction-only plus bundled code). That reduces risk from downloading remote installers, but the repository contains requirements files (notably a very large smyx_common requirements.txt) implying many dependencies would be needed to run the code. There is no automatic download URL or extract step. Because the skill includes many Python modules but no declared package installation step, running it may fail or require installing many third-party packages manually.

Credentials

The registry metadata declares no required environment variables or credentials, but the code reads and uses several environment/config locations: OPENCLAW_WORKSPACE (used to derive DB and file paths), OPENCLAW_SENDER_OPEN_ID / OPENCLAW_SENDER_USERNAME / FEISHU_OPEN_ID in ConstantEnum.init, and optional API keys in smyx_common config.yaml. The SKILL.md demands open-id and prefers reading api-key from skills/smyx_common/scripts/config.yaml (or workspace-level config) — that means sensitive identifiers or keys may be sourced from local config files or environment variables even though the manifest did not declare them. This mismatch is disproportionate and undocumented.

Persistence & Privilege

The skill writes persistent data: it will save uploaded attachments to an attachments directory and the DAO creates/uses an SQLite DB under a workspace data directory. The Base Enum YAML loader will create missing config.yaml files. The skill does not request always:true, but it does create and modify files in workspace-related paths and may persist analysis records locally. Those filesystem writes and local DB creation are not highlighted in the manifest or the SKILL.md's prohibition of reading local memory, creating a privilege/persistence inconsistency the user should be aware of.

What to consider before installing

Summary of practical concerns and steps before installing: - Data flows and privacy: This skill will accept facial images/videos and send them to remote APIs (base URLs found in smyx_common config point to lifeemergence.com domains). If you install it, highly sensitive biometric data will leave your machine — confirm you trust the remote service owner, their data retention and encryption practices, and that you have user consent. - Undeclared env/config access: The manifest claims no required env vars, but the code reads OPENCLAW_WORKSPACE, OPENCLAW_SENDER_OPEN_ID, OPENCLAW_SENDER_USERNAME, FEISHU_OPEN_ID and can read/write YAML config files (skills/smyx_common/scripts/config.yaml). Expect the skill to look at or create files under the workspace and to persist a local SQLite DB. If that surprises you, do not install. - Local persistence and writes: The skill will save attachments and may create a local DB under a workspace/data path and may create config YAML files if missing. Review those file locations and ensure they're acceptable for your environment (sandbox first). - Missing declaration of credentials: The code references API keys and has optional --api-key support but the registry lists no primary credential. Before use, determine what API endpoint and API key you must supply, who operates that endpoint, and how keys are stored and protected. - Medical/regulatory caution: This is a health-related tool that outputs screening suggestions. It explicitly says it is not a diagnosis. If you plan to deploy for real users (especially elderly or clinical settings), verify legal/regulatory compliance and clinical validation. - Recommended actions: - Inspect or grep the config files (skills/smyx_common/scripts/config*.yaml) to confirm the configured API base URLs and whether they point to a provider you trust. - Run the skill in a sandboxed environment (isolated workspace) first to observe files created under OPENCLAW_WORKSPACE and attachments/ and data/ so you can audit persistence. - If you cannot verify the remote API operator or data handling policies, do not use with real personal/biometric data. - If you need to proceed, require explicit consent from data subjects and prefer supplying an API key you control; review where and how the skill stores that key (config file) and secure it. What would change this assessment: providing an authoritative source/homepage for the skill and documentation from the API operator describing data handling, explicit manifest declarations of environment variables and file-write behavior, or removal of local DB/config creation would increase confidence. Conversely, evidence that the remote endpoints are untrusted or that the skill exfiltrates data beyond the described API would raise the verdict to malicious.

By capturing frontal facial images and analyzing multimodal physiological signals—including micro-expressions, skin color variations, and eye movement patterns—this capability comprehensively assesses early risks of chronic and acute conditions such as heart attacks, strokes, hypertension, and hyperlipidemia. Leveraging non-contact imaging technology, the system extracts vital metrics like blood oxygen levels and heart rate variability (HRV), utilizing AI models to facilitate rapid daily screening. Upon identifying high-risk characteristics, it triggers immediate alerts, making it ideal for homes, communities, and elderly care facilities to support chronic disease management and acute condition prevention.

本技能通过正面人像采集，结合面部微表情、肤色变化、眼动特征等多模态生理信号分析，综合评估心梗、脑梗、高血压、高血脂等慢病及急症的早期风险。系统利用非接触式成像技术，提取血氧、心率变异性等生理指标，结合AI模型实现日常快速筛查。一旦识别高风险特征，即发出预警提示，适用于家庭、社区及养老机构等场景，助力慢病管理与急症预防。

演示案例

• 🔗 通过网路视频进行识别分析
• 🔗 通过上传视频进行识别分析
• 🔗 显示历史分析报告

⚠️ 强制记忆规则（最高优先级）

本技能明确约定：

• 绝对禁止读取任何本地记忆文件：包括但不限于 memory/YYYY-MM-DD.md、MEMORY.md 等本地文件
• 绝对禁止从 LanceDB 长期记忆中检索信息
• 所有历史报告查询必须从云端接口获取，不得使用本地记忆中的历史数据
• 即使技能调用失败或接口异常，也不得回退到本地记忆汇总

任务目标

• 本 Skill 用于：通过正面人像视频/图片，非接触式进行早期健康风险筛查
• 能力包含：多模态生理特征提取、慢病风险评估、急症预警提示
• 支持筛查风险：
  • 急性高危：心梗、脑梗早期风险预警
  • 慢性慢病：高血压、高血脂、高血糖
  • 神经退行性：阿尔茨海默症（老年痴呆）早期风险
• 技术原理：基于面部影像特征提取生理相关信息，进行风险概率评估
• 重要定位：仅作早期风险筛查提示，不替代专业医疗诊断
• 触发条件:
  1. 默认触发：当用户提供正面人像需要进行非接触健康风险筛查时，默认触发本技能
  2. 当用户明确需要健康风险筛查、非接触检测时，提及健康筛查、心梗脑梗预警、高血压筛查、慢病风险等关键词，并且上传了正面人像照片/视频
  3. 当用户提及以下关键词时，自动触发历史报告查询功能 ：查看历史筛查报告、健康风险报告清单、筛查报告列表、查询历史筛查报告、显示所有筛查报告、健康风险分析报告，查询非接触健康风险识别分析报告
• 自动行为：
  1. 如果用户上传了附件或者照片/视频文件，则自动保存到技能目录下 attachments
  2. ⚠️ 强制数据获取规则（次高优先级）：如果用户触发任何历史报告查询关键词（如"查看所有筛查报告"、"显示所有风险报告"、" 查看历史报告"等），必须：
     • 直接使用 python -m scripts.contactless_health_risk_detection_analysis --list --open-id 参数调用 API 查询云端的历史报告数据
     • 严格禁止：从本地 memory 目录读取历史会话信息、严格禁止手动汇总本地记录中的报告、严格禁止从长期记忆中提取报告
     • 必须统一从云端接口获取最新完整数据，然后以 Markdown 表格格式输出结果

前置准备

• 依赖说明:scripts 脚本所需的依赖包及版本

  requests>=2.28.0
  

采集要求（获得准确结果的前提）

为了获得较准确的风险评估，请确保：

1. 正面完整面部，正对摄像头，距离 30-50 厘米
2. 光线充足均匀，避免强光直射和大面积阴影
3. 素颜最佳，避免浓妆、口罩、帽子、墨镜遮挡
4. 推荐采集：10-30 秒短视频，静态图片也支持分析

操作步骤

🔒 open-id 获取流程控制（强制执行，防止遗漏）

在执行非接触健康风险识别分析前，必须按以下优先级顺序获取 open-id：

第 1 步：【最高优先级】检查技能所在目录的配置文件（优先）
        路径：skills/smyx_common/scripts/config.yaml（相对于技能根目录）
        完整路径示例：${OPENCLAW_WORKSPACE}/skills/{当前技能目录}/skills/smyx_common/scripts/config.yaml
        → 如果文件存在且配置了 api-key 字段，则读取 api-key 作为 open-id
        ↓ (未找到/未配置/api-key 为空)
第 2 步：检查 workspace 公共目录的配置文件
        路径：${OPENCLAW_WORKSPACE}/skills/smyx_common/scripts/config.yaml
        → 如果文件存在且配置了 api-key 字段，则读取 api-key 作为 open-id
        ↓ (未找到/未配置)
第 3 步：检查用户是否在消息中明确提供了 open-id
        ↓ (未提供)
第 4 步：❗ 必须暂停执行，明确提示用户提供用户名或手机号作为 open-id

⚠️ 关键约束：

• 禁止自行假设,自行推导,自行生成 open-id 值（如 openclaw-control-ui、default、risk123、health456 等）
• 禁止跳过 open-id 验证直接调用 API
• 必须在获取到有效 open-id 后才能继续执行分析
• 如果用户拒绝提供 open-id，说明用途（用于保存和查询筛查报告记录），并询问是否继续

---

• 标准流程:
  1. 准备正面人像输入
     • 提供本地图片/短视频文件路径或网络 URL
     • 确保满足上述采集要求，获得更准确结果
  2. 获取 open-id（强制执行）
     • 按上述流程控制获取 open-id
     • 如无法获取，必须提示用户提供用户名或手机号
  3. 执行非接触健康风险识别分析
     • 调用 -m scripts.contactless_health_risk_detection_analysis 处理输入（必须在技能根目录下运行脚本）
     • 参数说明:
       • --input: 本地图片/视频文件路径（使用 multipart/form-data 方式上传）
       • --url: 网络图片/视频 URL 地址（API 服务自动下载）
       • --open-id: 当前用户的 open-id（必填，按上述流程获取）
       • --list: 显示历史非接触健康风险识别分析报告列表清单（可以输入起始日期参数过滤数据范围）
       • --api-key: API 访问密钥（可选）
       • --api-url: API 服务地址（可选，使用默认值）
       • --detail: 输出详细程度（basic/standard/json，默认 json）
       • --output: 结果输出文件路径（可选）
  4. 查看分析结果
     • 接收结构化的非接触健康风险识别分析报告
     • 包含：人像采集信息、各项疾病风险等级（低/中/高）、风险提示、建议就医指导

资源索引

必要脚本：见 scripts/contactless_health_risk_detection_analysis.py( 用途：调用 API 进行非接触健康风险识别分析，本地文件使用 multipart/form-data 方式上传，网络 URL 由 API 服务自动下载)

• 配置文件：见 scripts/config.py(用途：配置 API 地址、默认参数和格式限制)
• 领域参考：见 references/api_doc.md(何时读取：需要了解 API 接口详细规范和错误码时)

注意事项

• 仅在需要时读取参考文档，保持上下文简洁
• 支持格式：jpg/jpeg/png/mp4/avi/mov，视频推荐时长 10-30 秒，最大 100MB
• API 密钥可选，如果通过参数传入则必须确保调用鉴权成功，否则忽略鉴权
• ⚠️ 重要声明：本分析结果仅供早期风险筛查参考，不替代专业医疗诊断和检查，发现高风险请及时到医院就诊
• 禁止临时生成脚本，只能用技能本身的脚本
• 传入的网路地址参数，不需要下载本地，默认地址都是公网地址，api 服务会自动下载
• 当显示历史分析报告清单的时候，从数据 json 中提取字段 reportImageUrl 作为超链接地址，使用 Markdown 表格格式输出，包含" 报告名称"、"分析时间"、"高风险项数"、"风险等级"、"点击查看"五列，其中"报告名称"列使用健康风险筛查报告-{记录id}形式拼接, " 点击查看"列使用 [🔗 查看报告](reportImageUrl) 格式的超链接，用户点击即可直接跳转到对应的完整报告页面。
• 表格输出示例：

  报告名称	分析时间	高风险项数	整体风险	点击查看
  健康风险筛查报告 -20260328221000001	2026-03-28 22:10:00	1项(高血压)
  中风险	🔗 查看报告

使用示例

# 分析本地照片（以下只是示例，禁止直接使用openclaw-control-ui 作为 open-id）
python -m scripts.contactless_health_risk_detection_analysis --input /path/to/face.jpg --open-id openclaw-control-ui

# 分析网络视频（以下只是示例，禁止直接使用openclaw-control-ui 作为 open-id）
python -m scripts.contactless_health_risk_detection_analysis --url https://example.com/face.mp4 --open-id openclaw-control-ui

# 显示历史筛查报告/显示筛查报告清单列表/显示历史风险报告（自动触发关键词：查看历史筛查报告、历史报告、筛查报告清单等）
python -m scripts.contactless_health_risk_detection_analysis --list --open-id openclaw-control-ui

# 输出精简报告
python -m scripts.contactless_health_risk_detection_analysis --input face.jpg --open-id your-open-id --detail basic

# 保存结果到文件
python -m scripts.contactless_health_risk_detection_analysis --input face.jpg --open-id your-open-id --output result.json