A skill that teaches AI coding agents to debug systematically using runtime log evidence instead of guessing from code alone. It hooks into your existing agent (Claude Code, Codex, Cursor) and runs entirely on your machine.
How is this different from how agents normally debug?
Most agents jump to fixes by guessing from code alone, claiming 100% confidence but often failing. Debug Agent forces an evidence-based workflow: hypothesize, instrument with logs, reproduce, analyze runtime data, then fix only with proof.
What languages does it support?
JavaScript/TypeScript use a zero-dependency HTTP fetch to send logs. Python, Go, Rust, Java, C/C++, Ruby and other languages write NDJSON lines directly to a log file using standard file I/O.
Does it add dependencies to my project?
No. The logging server runs as a separate process via npx. Instrumentation uses a single fetch() call in JS/TS or native file I/O in other languages. Nothing is installed into your project.
Is there a hosted or enterprise version?
Coming soon. Email aiden@million.dev if you have questions or ideas.