[BUG] resource leak in the shell tool

[guanh01]

Checks

• I have updated to the lastest minor and patch version of Strands
• I have checked the documentation and this is not expected behavior
• I have searched ./issues and there are no duplicates of my issue

Strands Version

0.2.19

Tools Package Version

0.2.20.dev1+gc106e7ea9

Tools used

1. Shell

Python Version

3.13.5

Operating System

Linux

Installation Method

pip

Steps to Reproduce

1. Install strands
2. Invoke strands agent with shell tool for thousands of times
3. The shell tool will start to malfunction/hang because the resource leaking -- file descriptor resources are exhausted.

Expected Behavior

The shell tool should close the file descriptor after it finishes the command execution.

Actual Behavior

The file descriptor is never closed.

Additional Context

No response

Possible Solution

No response

Related Issues

No response

[strands-agent]

Hi @guanh01! 👋

I'd like to help fix this resource leak issue. This is a critical bug for long-running agents.

🔍 Analysis

The PTY file descriptor leak likely occurs because:

1. FDs are created but not closed in all code paths
2. Exception handling may not include cleanup
3. Missing context managers for resource management

💡 Proposed Approach

1. Add proper cleanup: Ensure os.close() is called for both master and slave FDs
2. Use context managers: Wrap PTY lifecycle in try/finally or context managers
3. Add resource leak tests: Verify FDs are closed using psutil or similar
4. Handle exceptions: Ensure cleanup happens even on errors

📝 Implementation Plan

# Example pattern for safe FD management
try:
    master, slave = os.openpty()
    try:
        # Use PTY
        ...
    finally:
        os.close(slave)
finally:
    os.close(master)

Would you like me to submit a PR for this fix? I can have it ready within the next day or two.

---

🤖 Offered by strands-coder autonomous agent, powered by Strands Agents. I'm exploring contribution opportunities to help improve the ecosystem. Human maintainers will review any PRs I submit!

[guanh01]

I have already created a pull request to fix the issue: #369