<!-- # COPYRIGHT NOTICE # This file is part of the "Universal Biomedical Skills" project. # Copyright (c) 2026 MD BABU MIA, PhD <md.babu.mia@mssm.edu> # All Rights Reserved. # # This code is proprietary and confidential. # Unauthorized copying of this file, via any medium is strictly prohibited. # # Provenance: Authenticated by MD BABU MIA -->

---

name: 'pandas-best-practices' description: 'Standards for efficient, readable, and performant data manipulation using Python''s Pandas library.' measurable_outcome: Execute skill workflow successfully with valid output within 15 minutes. allowed-tools:

• read_file
• run_shell_command
• write_file

---

Pandas Best Practices

This skill provides guidelines for working with tabular data in Python. It focuses on vectorization, memory management, and method chaining to write "Modern Pandas" code.

When to Use This Skill

• Data Cleaning: Preprocessing clinical or genomic datasets.
• Analysis: Performing aggregations, merges, or statistical summaries.
• Performance: Optimizing slow-running scripts that process large CSVs/DataFrames.

Core Capabilities

1. Vectorization: Replacing

   for

   loops with vectorized array operations.
2. Method Chaining: Writing readable, fluent data transformation pipelines.
3. Memory Optimization: Using appropriate dtypes (Categoricals, Nullable Ints) to reduce RAM usage.
4. Modern Indexing: Using

   .loc

   and

   .iloc

   correctly; avoiding

   SettingWithCopyWarning

   .

Workflow

1. Inspect Data: Check

   df.info()

   and

   df.head()

   .
2. Define Pipeline: Plan transformations (filter -> group -> aggregate).
3. Implement Chain: Write the logic as a chain of methods.
4. Optimize: Check for loops or

   apply

   calls that can be vectorized.

Example Usage

User: "Calculate the mean age by patient group, but exclude patients with missing IDs."

Agent Action:

1. Reads

   references/rules.md

   .
2. Generates:

   result = (
       df
       .dropna(subset=['patient_id'])
       .groupby('patient_group')['age']
       .mean()
       .reset_index()
   )
   

<!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->