\r \r

When to Use This Skill\r

\r Activate this skill when any of the following conditions are met:\r \r

• The user sends a PDF link of a checkup report and asks to analyze or interpret it\r
• The user provides local files (PDF or images) of a checkup report and asks to analyze them\r
• The user says things like "help me analyze my checkup report", "look at my health report", "interpret this medical report", or similar requests (in any language)\r
• The user mentions checkup report analysis needs and provides a PDF URL or local file paths\r \r Note: This skill supports both publicly accessible PDF URLs and local file uploads (PDF and images). The most common usage scenario is uploading local files via binary stream (--file), as users typically have report files saved on their device rather than hosted at a public URL.\r \r How to choose (mutually exclusive — pick one):\r
• If the user provides a local file path (e.g. /home/user/report.pdf, C:\Users\...\report.pdf, or relative path) → use --file (binary upload). This is the primary and most common usage.\r
• If the user provides a URL (starts with http:// or https://) → use --url\r
• Do NOT use --url and --file together in one command\r \r ---\r \r

⚠️ Privacy Warning\r

\r IMPORTANT - READ BEFORE INSTALLING:\r \r This skill uploads your checkup report to WiseDiag's cloud servers (via PDF URL or direct binary file upload) for AI-powered health analysis.\r \r Do NOT use with sensitive or confidential medical documents unless:\r

• You trust WiseDiag's data handling policies\r
• You accept that file contents will be transmitted and processed remotely\r \r For sensitive documents, use offline/local checkup report analysis tools instead.\r \r ---\r \r

WiseAnalyze Skill (powered by WiseDiag)\r

\r An AI-powered medical checkup report analysis tool. Submit a checkup report (via local file upload or publicly accessible PDF URL), and the system will return a structured health interpretation including abnormal items, clinical explanations, lifestyle advice, and personalized recommendations.\r \r The analysis uses an async task model with 4 separate sub-commands. You (the Agent) must call each step and handle polling yourself.\r \r

Installation\r

\r

pip install -r requirements.txt\r
```\r
\r
## 🔑 API Key Setup (Required)\r
\r
**Get your API key:** 👉 [https://console.wisediag.com/apiKeyManage](https://s.wisediag.com/xsu9x0jq)\r
\r
The API key MUST be set as an environment variable. The script reads it automatically.\r
\r
```bash\r
export WISEDIAG_API_KEY=your_api_key\r
```\r
\r
## How to Analyze a Report (Step-by-Step)\r
\r
**NEVER call any API or HTTP endpoint directly. ONLY use the script sub-commands below.**\r
\r
**⚠️ IMPORTANT: Each command makes HTTP requests to an external API and may take 10-30 seconds to complete. Do NOT kill or interrupt a running command — wait for it to finish. The script prints progress messages so you can see it is still working.**\r
\r
The workflow has 4 steps. You MUST follow them in order.\r
\r
---\r
\r
### Step 1: Submit local files or PDF URL\r
\r
There are two ways to submit — pick one (mutually exclusive):\r
\r
**Option A: Upload local files (recommended)** — when user provides local file paths\r
\r
```bash\r
cd scripts\r
python3 wise_analyze.py submit -f "/path/to/checkup_report.pdf"\r
python3 wise_analyze.py submit -f "/path/to/report.pdf" -f "/path/to/scan.jpg"\r
```\r
\r
**Option B: Submit via URL** — when user provides a public HTTP/HTTPS link\r
\r
```bash\r
cd scripts\r
python3 wise_analyze.py submit -u "https://example.com/checkup_report.pdf"\r
```\r
\r
With optional health questionnaire (recommended for better results):\r
\r
```bash\r
python3 wise_analyze.py submit -f "/path/to/report.pdf" -q "occasional chest tightness, family history of hypertension, poor sleep quality"\r
```\r
\r
With optional member ID to link to a health profile:\r
\r
```bash\r
python3 wise_analyze.py submit -f "/path/to/report.pdf" -m "HR1017911158367870xxxx"\r
```\r
\r
**Output:** The script prints `TASK_ID=\x3Cid>`. Save this task ID for the next steps.\r
\r
**How to choose (mutually exclusive — pick one):**\r
- User gives a local file path (e.g. `/home/user/report.pdf`, `C:\Users\...\report.pdf`) → use `-f`\r
- User gives a URL (starts with `http://` or `https://`) → use `-u`\r
- Do NOT use `-u` and `-f` together in one command\r
\r
---\r
\r
### Step 2: Start interpretation\r
\r
```bash\r
cd scripts\r
python3 wise_analyze.py start -t "\x3Ctask_id>"\r
```\r
\r
**Output:** Prints confirmation that interpretation has started.\r
\r
---\r
\r
### Step 3: Poll for progress (YOU must repeat this)\r
\r
**⚠️ CRITICAL: You MUST NOT skip this step. Do NOT jump directly to Step 4.**\r
\r
```bash\r
cd scripts\r
python3 wise_analyze.py query -t "\x3Ctask_id>"\r
```\r
\r
**Output:** The script prints machine-readable status:\r
\r
```\r
STATUS=processing\r
PROGRESS=50\r
```\r
\r
**Polling rules:**\r
- If `STATUS=queued` or `STATUS=processing` → **tell the user the current progress percentage**, then wait and poll again using **dynamic intervals based on progress**:\r
  - `PROGRESS` 0–30%: wait **10 seconds** before next query\r
  - `PROGRESS` 30–70%: wait **8 seconds** before next query\r
  - `PROGRESS` 70%+: wait **5 seconds** before next query\r
- If `STATUS=finish` → proceed to Step 4\r
- If `STATUS=failed` → the task failed, inform the user\r
\r
**You MUST:**\r
1. Keep calling the query command until STATUS is `finish` or `failed`. Do NOT give up after one check.\r
2. **Report each polling result to the user** so they can see the analysis progress (e.g. "Analysis progress: 50%..."). Do NOT silently poll without updating the user.\r
3. Never skip Step 3 and go directly to Step 4. The `save` command will fail if the task is not finished.\r
\r
---\r
\r
### Step 4: Save the result\r
\r
```bash\r
cd scripts\r
python3 wise_analyze.py save -t "\x3Ctask_id>" -n "report_2025"\r
```\r
\r
**Output:** The command saves the report and prints `REPORT_PATH=\x3Cpath>`.\r
\r
If `-n` is not provided, the task ID is used as the filename.\r
\r
**⚠️ IMPORTANT — How to present the result to the user:**\r
\r
After the save command completes, you MUST:\r
1. Read the saved `.md` file using the path printed in `REPORT_PATH=`\r
2. Output a brief one-sentence introduction (e.g. "以下是您的体检报告分析结果：")\r
3. Output the full report content **directly** as-is (rendered Markdown) — do NOT summarize, condense, reorganize, or rephrase any part of it\r
4. After the full report, append a short "关键提醒" section: extract 2-3 most critical items from the report that need immediate attention, keep each point to one sentence\r
5. Do NOT omit or shorten the report body itself — the full report must be shown in its entirety between the intro and the closing reminder\r
\r
---\r
\r
## Sub-commands Reference\r
\r
### submit\r
\r
| Flag | Description |\r
|------|-------------|\r
| `-u, --url` | Publicly accessible URL of the PDF checkup report (optional, requires --url or --file) |\r
| `-f, --file` | Local file path to upload, PDF or image (optional, repeatable for multiple files) |\r
| `-q, --questionnaire` | Health questionnaire text: symptoms, family history, lifestyle (optional but recommended) |\r
| `-m, --member-id` | Health profile member ID to link report to existing profile (optional) |\r
\r
### start\r
\r
| Flag | Description |\r
|------|-------------|\r
| `-t, --task-id` | Task ID returned from the submit step (required) |\r
\r
### query\r
\r
| Flag | Description |\r
|------|-------------|\r
| `-t, --task-id` | Task ID to check status (required) |\r
\r
### save\r
\r
| Flag | Description |\r
|------|-------------|\r
| `-t, --task-id` | Task ID to save result (required) |\r
| `-n, --name` | Output filename stem (default: taskId) |\r
| `-o, --output` | Output directory (default: current working directory) |\r
\r
## Data Privacy\r
\r
**What happens to your files:**\r
1. When using `--url`: The PDF URL is sent to WiseDiag's analysis API, downloaded and processed on their servers\r
2. When using `--file`: Local files are uploaded directly to WiseDiag's servers via multipart/form-data\r
3. AI-powered health interpretation results are returned to you\r
4. Files are not permanently stored on WiseDiag servers\r
\r
**For sensitive documents, use offline/local checkup report analysis tools instead.**\r
\r
## License\r
\r
MIT\r