stas00stas00Communityยท1 Agent Skills Included

the-art-of-debugging

Debug crashes, hangs, and memory errors in Python and PyTorch fast

Diagnoses crashes, segfaults, hangs, CUDA out-of-memory errors, NaN values, and slow code in Unix, Python, and PyTorch programs. Replaces guesswork with a proven loop: reproduce the bug, shrink the payload, localize the cause, and verify the fix. Provides copy-paste recipes for gdb, strace, py-spy, and profilers so your AI agent finds root causes in minutes.
npx skills add stas00/the-art-of-debugging --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install The Art of Debugging?โ–ผ

Run `npx skills add stas00/the-art-of-debugging --all -g -y` in your terminal to install this debugging skill globally for your AI agent.

How to fix a CUDA out of memory error?โ–ผ

This skill teaches your agent to distinguish forward vs backward OOM, tune memory fragmentation settings, and trace exactly which tensors consumed GPU memory.

How to find why my Python program hangs?โ–ผ

It uses py-spy to dump live stack traces from a stuck process without changing any code, revealing the exact line where execution is blocked.

Can AI debug multi-GPU training deadlocks?โ–ผ

Yes. The skill guides your agent to test node communication, dump every rank's stack at once, and spot mismatched collective operations causing the deadlock.

Does this work with Claude Code and Cursor?โ–ผ

Yes. It follows the universal SKILL.md standard and runs in Claude Code, Cursor, Codex, and other compatible AI coding agents.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’