Show HN: Contextrot– Analyze where Claude Code degrades from its own session log

Category: observability

Tags: llm-agent, context-rot, claude-code, observability, python

Score: 7.3/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 7)

Contextrot analyzes Claude Code session logs to detect when and why an LLM coding agent's performance degrades as its context window fills, providing a per-user diagnostic rather than a generic benchmark. It extracts five failure signals (edit failures, retry loops, re-reads, self-corrections, tool errors) and correlates them with context fill, offering actionable prescriptions. This is interesting because it fills a gap in understanding real-world agent behavior, is completely local and privacy-respecting, and can say "you're fine" when no rot is detected, building trust in its warnings.

Target audience: backend devs, devops, AI developers

Repository: https://github.com/Priyanshu-byte-coder/contextrot · Python · MIT · 10 stars

View on Hacker News