CtrlK
BlogDocsLog inGet started
Tessl Logo

sandbox-exec-fallback

Fallback pattern for executing Python code when execute_code_sandbox fails by writing to file and running via shell

48

Quality

50%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./benchmarks/gdpval/skills/sandbox-exec-fallback/SKILL.md
SKILL.md
Quality
Evals
Security

Sandbox Execution Fallback

When to Use

Use this skill when execute_code_sandbox fails during Python code execution, particularly for tasks involving:

  • Spreadsheet generation (xlsx, csv)
  • Data processing and transformations
  • File output operations
  • Complex computations requiring libraries

Recovery Pattern

When execute_code_sandbox returns an error or fails to produce expected output:

Step 1: Write Python Code to File

Use write_file to save your Python script to a file:

# Use write_file tool with:
path: "script.py"
content: <your complete Python code as a string>

Step 2: Execute via Shell

Use run_shell to execute the script:

# Use run_shell tool with:
command: "python3 script.py"
timeout: <appropriate timeout, e.g., 300 for long-running tasks>

Step 3: Verify Output

Check the shell output for:

  • Success messages or printed results
  • Error messages (if any, diagnose and fix the script)
  • File creation confirmations

Step 4: Access Generated Files

If the script creates output files (e.g., spreadsheets), they will be in the current working directory. Use list_dir to confirm, then read_file to access if needed.

Why This Works

  • Same Python Environment: python3 in the shell uses the same environment as the sandbox
  • More Robust: Shell execution handles long-running or memory-intensive tasks better
  • Debuggable: Errors are captured in stdout/stderr for easier diagnosis
  • Identical Results: Produces the same output as sandbox execution when successful

Example: Spreadsheet Generation

Original code that failed in execute_code_sandbox:

import pandas as pd

data = {'A': [1, 2, 3], 'B': [4, 5, 6]}
df = pd.DataFrame(data)
df.to_excel('output.xlsx', index=False)
print("File created successfully")

Recovery Execution:

  1. write_file with path="generate_spreadsheet.py" and the code above as content
  2. run_shell with command="python3 generate_spreadsheet.py"
  3. Verify output.xlsx was created with list_dir

Tips

  • Increase timeout for complex operations (default 30s may be insufficient)
  • Add explicit print statements for debugging
  • Handle exceptions in your script and print error messages
  • Clean up temporary script files after successful execution if needed
Repository
HKUDS/OpenSpace
Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.