Feedback: Incredible experience using Claude Code to contribute to sandbox-runtime

Resolved 💬 3 comments Opened Dec 22, 2025 by abe238 Closed Feb 14, 2026

Experience Summary

I just had an incredible experience using Claude Code to contribute to the @anthropic-ai/sandbox-runtime project. The workflow was seamless and deeply impressive.

What Happened

  1. Problem Discovery: I encountered a setRawMode EPERM error when trying to run Claude Code inside srt (sandbox-runtime)
  1. Deep Analysis: Claude Code helped me:
  • Explore the sandbox-runtime codebase
  • Understand the macOS Seatbelt profile generation
  • Discover that allowPty functionality already existed but wasn't exposed to users
  1. Solution Design: Using sequential thinking, we:
  • Proposed 3 different solutions
  • Conducted adversarial debate on each approach
  • Selected the best option (explicit --tty CLI flag)
  1. Implementation: Claude Code then:
  • Forked the repository
  • Created a feature branch
  • Implemented the --tty flag (~30 lines of code)
  • Updated README documentation
  • Built and tested the changes
  • Created a well-documented PR with detailed technical analysis

Links

What Made This Special

  1. End-to-end ownership: From problem identification → research → solution design → implementation → PR submission
  2. Deep reasoning: The adversarial analysis of solutions felt like pair programming with a senior engineer
  3. Humble contribution: The goal was adding genuine value to open source, not just generating code
  4. Documentation: The PR includes extensive reasoning so future maintainers understand the "why"

The Irony

Claude Code helped fix a limitation in the very sandbox that wraps Claude Code. 🤯

Thank you for building such an incredible tool. This experience genuinely felt like the future of software development.

---

🤖 This feedback was written collaboratively with Claude Code

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗