Pretend-Sick 🤒

帮请病假 - AI推测合理病症、生成请假话术、指导开证明、管理生病状态

Security Scan

The skill's purpose (helping craft sick-leave messages and track illness) is plausible, but the runtime instructions reference missing scripts and an external image-model API without declaring required code or credentials, making the package incoherent and incomplete — investigate before installing or trusting it with photos/health data.

Purpose & Capability

The name/description align with the listed features (symptom inference, leave templates, state tracking). However, the SKILL.md claims image recognition via scripts/img-analyze.py (GLM-4.5V API) and a CLI 'sick-buddy', yet the package contains no executable scripts or those files. That mismatch (declared capabilities but missing implementation/dependencies) is disproportionate and unexplained.

Instruction Scope

Instructions ask the agent to accept photos and run an image-analysis script and to WebSearch for local procedures. They also specify local storage paths (~/.openclaw/workspace/sick-buddy-data/*.json) for personal health data. While storing state is within scope, calling an external model/API for photo analysis is outside the declared environment (no script or API key), and the agent could end up sending sensitive images to remote services unless that behavior is explicitly controlled.

Install Mechanism

There is no install spec (instruction-only), which is low risk for arbitrary installs. README suggests a git repo URL, but that is informational only — the registry package itself contains no installers or downloads.

Credentials

requires.env is empty, yet SKILL.md references using the GLM-4.5V API for image analysis. That API would normally require credentials and potentially an endpoint; no env vars or credential requirements are declared. The skill also references no other external credentials, which is appropriate, but the missing declaration for the image/model capability is a notable omission.

Persistence & Privilege

The skill writes user data to two files under the user's OpenClaw workspace (~/.openclaw/workspace/sick-buddy-data). This local persistence is expected for a stateful assistant and the skill does not request elevated or system-wide privileges, nor does it force global inclusion (always:false).

What to consider before installing

Before installing or using this skill: 1) Treat it as incomplete — SKILL.md asks to run scripts/img-analyze.py and a CLI 'sick-buddy' but those scripts are not present in the package; ask the author for the repository or the missing files and inspect them. 2) Verify how image analysis is implemented and where photos will be sent — if it uses GLM-4.5V or another remote model, confirm which endpoint and which credentials are used; do not provide sensitive photos unless you trust the service and encryption. 3) Be aware the skill stores personal health data in ~/.openclaw/workspace/sick-buddy-data; decide whether that storage is acceptable (unencrypted local files) and periodically clear or back up as you see fit. 4) Consider ethical/policy implications — the tool can help craft excuses for absence; ensure its use complies with your employer/school rules. 5) If you need to proceed, request the missing code and credentials, review them locally for network calls or exfiltration, and only enable the skill once the implementation matches the declared behavior.

Pretend-Sick 🤒 — AI生病管理助手

帮生病的人安心休息：推测病症、了解病程发展、生成请假消息、每日提醒注意事项。

设计原则

1. 少问多做 — 用户说了需求直接给方案，不连续追问
2. 天数匹配病症 — 请1天=轻微，3天=中等，5天+=严重
3. 优先社区医院 — 社康更容易开证明，更快
4. 先开证明再请假 — 建议先拿证明再提交
5. 灵活补充 — 需要更多天就补充症状

需求澄清（动手前快速确认）

用户说了需求后，快速确认这3点（缺哪个问哪个，不挨个问）：

#	问题	为什么	默认值
1	请几天？	决定病症严重程度	用户说的
2	哪里不舒服？	推测病症	用户说的
3	公司/学校用什么请假系统？	生成对应话术	飞书

如果3个信息都有了，直接给方案，不追问。

核心功能

1. 症状推测

用户用大白话描述 + 拍照，AI推测可能的病症，给出多种可能性+各自休息天数。 结合城市+季节+当地疫情辅助推测。

2. 病程管理

• 个人化时间线（第几天会怎样）
• 多病交叉分析（症状叠加、药物冲突）
• 每日提醒（今天可能怎样、该怎么做）
• 就医提醒（危险信号）

3. 请假执行

• 问清公司/学校+请假系统（钉钉/飞书/企微/纸质）
• WebSearch查具体请假流程
• 生成请假话术（正式/随意）
• 帮找快速获取证明的渠道（社区医院/线上问诊/先请后补）
• 生成交接消息
• 确认清单，不让用户漏任何一步

4. 动态案例

根据用户实际情况（城市+公司+症状）实时生成"跟你类似的人怎么处理的"。 优先搜真实经验贴，搜不到用知识库组合。

5. 图片识别

支持拍照识别：喉咙红肿、皮肤红疹、体温计读数、药品识别、处方单读取。 调用 scripts/img-analyze.py（GLM-4.5V API）。

6. 证明获取指导

• 社区医院：最快最便宜，随到随开
• 线上问诊：微医/好大夫/丁香医生，30分钟拿到
• 先请后补：查公司政策，帮写说明
• 告诉证明长什么样、要盖什么章、带什么证件

数据存储

• ~/.openclaw/workspace/sick-buddy-data/profile.json — 当前病症、开始日期
• ~/.openclaw/workspace/sick-buddy-data/daily-log.json — 每日状态记录

CLI工具

sick-buddy start 感冒        # 建档
sick-buddy symptoms 嗓子疼   # 添加症状
sick-buddy set city 杭州     # 设置城市
sick-buddy set workplace 阿里 # 设置公司
sick-buddy confirm 普通感冒  # 确认病症
sick-buddy status            # 查看状态
sick-buddy log 今天好多了     # 记录状态
sick-buddy history           # 历史记录
sick-buddy clear             # 清除记录

安全声明

⚠️ 本工具不提供医疗诊断，所有信息仅供自我护理参考。严重症状请及时就医。

知识库

• knowledge/diseases.md — 常见疾病（感冒/流感/肠胃炎/偏头痛/咽炎/过敏/腰肌劳损）+ 多病交叉注意事项
• knowledge/leave-templates.md — 请假话术模板（公司/学校/正式/随意）