Skip to content

Code Execution Tools

OpenJarvis includes 4 code tools for running Python, executing shell commands, checking Python syntax, and running pytest test suites.

Tools

run_python

Execute Python code in a subprocess and return its output.

Parameters:

Parameter Type Default Description
code string required Python code to execute
timeout integer 10 Timeout in seconds

Returns: Combined stdout and stderr output (capped at 4KB).

Example prompts: 

> Run this Python code and show me the output: print([x**2 for x in range(10)])
> Execute: import json; print(json.dumps({"key": "value"}, indent=2))
> Calculate the Fibonacci sequence up to 100 using Python

run_shell

Execute a shell command and return its output.

Parameters:

Parameter Type Default Description
command string required Shell command to execute
timeout integer 15 Timeout in seconds

Returns: Combined stdout and stderr output (capped at 4KB).

Example prompts: 

> Run: ls -la | grep .py
> Execute: git log --oneline -10
> Check the disk space: df -h

lint_python

Check Python code for syntax errors without executing it.

Parameters:

Parameter Type Required Description
code string yes Python code to check

Returns: "Syntax OK" or an error message with line number.

Example prompts: 

> Check if this Python code has any syntax errors: [paste code]
> Validate this Python snippet before running it

run_pytest

Run pytest on a test suite or specific test file and return execution output with pass/fail counts.

Parameters:

Parameter Type Default Description
test_path string "" Optional path to specific test file or directory
args string "" Optional additional pytest arguments (e.g. "-k test_editor -v")

Returns: Output string including exit code, test pass/fail counts, and failure tracebacks.

Usage Examples

Running Python Subprocesses

oj> Calculate the sum of squares from 1 to 100 using Python

  ↳ routing: generalist → code
  ⚙ tool: run_python {"code": "print(sum(i**2 for i in range(1, 101)))"}
    → 338350

The sum of squares from 1 to 100 is **338,350**.

Checking Generated Python Code

oj> Check if this Python function has any syntax issues: def foo(x): return x + 1

  ↳ routing: generalist → code
  ⚙ tool: lint_python {"code": "def foo(x):\n    return x + 1"}
    → "Syntax OK"

The snippet contains valid Python syntax without errors.

Security & Timeout Guardrails

Code execution tools run directly in isolated subprocesses using the active user's permissions:

  • Subprocess Isolation: Subprocesses are spawned per-execution without shell string injection.
  • Strict Timeouts: 10 seconds for run_python, 15 seconds for run_shell.
  • Buffer Cap: Standard output and error streams are truncated to 4KB.
  • Static Linting (lint_python): Uses Python's internal AST parser to validate syntax safely without executing code.

See Also