---

name: python_exec description: Execute Python code in an isolated environment for calculations, file generation, and data processing version: 1.0.0 author: Architecture Team license: MIT dependencies:

• python>=3.11
• python-docx>=1.1.0
• openpyxl>=3.1.0
• pandas>=2.0.0
• matplotlib>=3.8.0
• reportlab>=4.0.0
• Pillow>=10.0.0 tags:
• execution
• python
• code
• calculations
• document-generation
• data-processing

---

Python Execution Skill

Execute Python code in an isolated subprocess environment with timeout protection. This skill is essential for:

• Document Generation: Create Word (.docx), Excel (.xlsx), PowerPoint (.pptx), and PDF files
• Calculations: Perform mathematical operations, statistical analysis, and financial computations
• Data Processing: Transform, aggregate, and analyze data from various sources
• Research Steps: Intermediate computations during multi-step research workflows

When to Use

Document Generation

Use this skill when you need to create binary document formats that cannot be generated with simple file writes:

• Word documents with formatted tables, headers, styles
• Excel spreadsheets with formulas, charts, multiple sheets
• PowerPoint presentations with slides, images, layouts
• PDF reports with styling, tables, embedded images

Calculations & Analysis

Use this skill for computational tasks during query processing:

• Mathematical calculations (percentages, averages, sums, ratios)
• Financial metrics (CAGR, growth rates, margin calculations)
• Statistical analysis (mean, median, standard deviation)
• Date/time calculations and formatting
• Aggregating data from multiple sources

Deep Research Steps

Use this skill as an intermediate step in research workflows:

• Process and validate extracted data from documents
• Compute comparisons between entities (e.g., ESG scores)
• Generate summary statistics from research findings
• Transform extracted JSON/CSV data into structured outputs
• Merge data from multiple document extractions

Visualization

Use this skill to create visual representations of data:

• Charts and graphs with matplotlib or plotly
• Data visualizations for reports
• Infographics from analyzed data

When NOT to Use

• Writing plain text or markdown files (use file_write)
• Reading files (use file_read or document_qa)
• Searching files (use file_grep or file_glob)
• Running shell commands (use bash_exec)
• Simple string manipulation
• Just displaying text output

Input Schema

Input Example

Copy{
  "code": "import pandas as pd\nfrom openpyxl import Workbook\n\n# Create Excel with data\nwb = Workbook()\nws = wb.active\nws['A1'] = 'Company'\nws['B1'] = 'ESG Score'\nws['A2'] = 'Tata Steel'\nws['B2'] = 85.5\nwb.save('/tmp/esg_report.xlsx')\nprint('Created: /tmp/esg_report.xlsx')",
  "timeout_seconds": 60
}

Output Schema

Output Example

Copy{
  "output": "Created: /tmp/esg_report.xlsx",
  "error": null,
  "return_value": null,
  "execution_time_ms": 245.67,
  "success": true
}

Examples

Create a Word Document

Copyfrom docx import Document
from docx.shared import Inches, Pt
from docx.enum.text import WD_ALIGN_PARAGRAPH

doc = Document()

# Add title
title = doc.add_heading('ESG Report: Tata Steel', 0)
title.alignment = WD_ALIGN_PARAGRAPH.CENTER

# Add summary paragraph
doc.add_paragraph('This report summarizes the key ESG performance indicators...')

# Add table
table = doc.add_table(rows=4, cols=3)
table.style = 'Table Grid'
headers = table.rows[0].cells
headers[0].text = 'Category'
headers[1].text = 'Score'
headers[2].text = 'Trend'

data = [
    ('Environmental', '82.5', 'Improving'),
    ('Social', '78.3', 'Stable'),
    ('Governance', '91.2', 'Improving'),
]

for i, (cat, score, trend) in enumerate(data, 1):
    row = table.rows[i].cells
    row[0].text = cat
    row[1].text = score
    row[2].text = trend

doc.save('/Users/output/esg_report.docx')
print('Document created successfully')

Calculate Financial Metrics

Copyimport json

# Data extracted from documents
data = {
    'revenue_2022': 150000000,
    'revenue_2023': 175000000,
    'costs_2023': 140000000,
    'total_assets': 500000000,
    'total_liabilities': 200000000,
}

# Calculate metrics
revenue_growth = ((data['revenue_2023'] - data['revenue_2022']) / data['revenue_2022']) * 100
profit_margin = ((data['revenue_2023'] - data['costs_2023']) / data['revenue_2023']) * 100
debt_ratio = (data['total_liabilities'] / data['total_assets']) * 100

results = {
    'revenue_growth_pct': round(revenue_growth, 2),
    'profit_margin_pct': round(profit_margin, 2),
    'debt_ratio_pct': round(debt_ratio, 2),
}

print(json.dumps(results, indent=2))

Generate Chart from Data

Copyimport matplotlib.pyplot as plt

companies = ['Tata Steel', 'JSW Steel', 'SAIL', 'Hindalco']
esg_scores = [85.5, 78.2, 72.1, 81.3]

plt.figure(figsize=(10, 6))
bars = plt.bar(companies, esg_scores, color=['#2E86AB', '#A23B72', '#F18F01', '#C73E1D'])
plt.xlabel('Company')
plt.ylabel('ESG Score')
plt.title('ESG Score Comparison - Indian Steel Companies')
plt.ylim(0, 100)

for bar, score in zip(bars, esg_scores):
    plt.text(bar.get_x() + bar.get_width()/2, bar.get_height() + 1,
             f'{score}', ha='center', va='bottom', fontweight='bold')

plt.tight_layout()
plt.savefig('/tmp/esg_comparison.png', dpi=150)
print('Chart saved to /tmp/esg_comparison.png')

Aggregate Research Data

Copyimport json

# Extracted data from multiple documents
tata_data = {'env': 82, 'social': 78, 'gov': 91}
jsw_data = {'env': 75, 'social': 80, 'gov': 85}
sail_data = {'env': 68, 'social': 72, 'gov': 78}

# Aggregate and compare
companies = {
    'Tata Steel': tata_data,
    'JSW Steel': jsw_data,
    'SAIL': sail_data,
}

summary = []
for name, scores in companies.items():
    avg = sum(scores.values()) / len(scores)
    summary.append({
        'company': name,
        'environmental': scores['env'],
        'social': scores['social'],
        'governance': scores['gov'],
        'overall_avg': round(avg, 1),
    })

# Sort by overall average
summary.sort(key=lambda x: x['overall_avg'], reverse=True)

print('ESG Rankings:')
for i, company in enumerate(summary, 1):
    print(f"{i}. {company['company']}: {company['overall_avg']}")

print('\nDetailed Results:')
print(json.dumps(summary, indent=2))

Available Libraries

The following libraries are pre-installed and available:

Safety & Limitations

• Timeout: Maximum execution time is 300 seconds (5 minutes)
• Output Limit: Output is truncated at 30,000 characters
• Isolation: Code runs in a subprocess for isolation
• Approval: This skill requires policy approval before execution
• No Network: External network calls should use dedicated skills

Best Practices

1. Print outputs: Always print results so they appear in the output
2. Save files to accessible paths: Use paths the user can access
3. Handle errors gracefully: Use try/except for robust execution
4. Use appropriate libraries: Choose the right library for the task
5. Keep code focused: One clear objective per execution