Show HN: Amidumb – scores how you direct Claude Code and Codex, from local logs

Category: devtools

Tags: ai-coding, developer-tools, transcript-analysis, privacy, scoring

Score: 6.3/10 (Innovation: 6, Technical: 6, Documentation: 8, Utility: 5)

amidumb analyzes local Claude Code and Codex transcripts to score how effectively a developer directs an AI coding agent, using a fixed rubric for objectivity and an LLM only for playful roasts. It stands out for its privacy-first design, contextual prompt classification, and detailed transparency about its scoring methodology and limitations.

Target audience: AI-assisted developers, engineering managers, and teams adopting Claude Code or Codex

Repository: https://github.com/jjinendra3/amidumb · TypeScript · MIT

View on Hacker News