elisaterumi-ai/clinical-agent-skills/classification

When classifying clinical text, follow these steps:

1. Understand the classification task

• Identify the classification dimension from the context (e.g., severity, urgency, type)
• Use only the categories provided or implied

2. Extract relevant signals

Identify information that supports classification:

• Symptoms and their severity
• Clinical findings
• Diagnoses
• Risk indicators
• Temporal aspects (acute vs chronic)

Only use explicitly stated information.

3. Assign a label

Classify the case into one of the predefined categories.

Examples (depending on context):

• Severity: MILD / MODERATE / SEVERE
• Urgency: LOW / MEDIUM / HIGH
• Type: ACUTE / CHRONIC

Do not create new categories.

4. Justify the classification

Provide a short rationale based on:

• Key symptoms or findings
• Clinical indicators present in the text

Keep justification concise and grounded.

5. Handle uncertainty

• If information is insufficient, state uncertainty
• Do not guess or assume missing data

6. Output format

Return the result as structured text:

Label: ...
 
Rationale:
- ...
- ...

7. Constraints

• Only use categories defined in the context
• Do not hallucinate clinical information
• Do not infer missing data
• Do not introduce external knowledge

Do not provide medical advice or recommendations