← Work

2026 · Author

Auto-Debug Agent

An agent that writes Python, runs it, and fixes its own mistakes.

  • Python
  • LLM tool-use
  • Sandboxed execution
Source

A compact demonstration of self-correcting autonomy: the agent drafts code, runs it in a sandbox, reads the traceback when it fails, and revises — looping until the program works or it gives up gracefully.

A small, readable codebase that makes the write → run → observe → repair cycle obvious, without hiding it behind a framework.