context-rot

To audit AI context files for stale or redundant instructions, scan your CLAUDE.md and recent conversation transcripts to detect friction signals like user corrections and repeated clarifications, generating a scored health report with prioritized fixes.

Context rot is the accumulation of stale, redundant, or missing instructions in AI context files and transcripts that degrades performance and increases operational costs by causing costly misunderstandings and repeated clarifications.

Fix missing or bloated CLAUDE.md instructions by reviewing the markdown documentation for anti-patterns, redundancy, and ghost sections, then applying actionable fixes from a prioritized queue based on transcript friction analysis.

Yes, you can detect conversation friction signals by scanning recent transcripts for user corrections, repeated clarifications, and policy overrides, which indicate underlying issues in your foundational AI context instructions.

Run a context health audit before a major project kickoff to ensure your AI agent's foundational instructions are clear, concise, and up-to-date, preventing costly misunderstandings and improving overall operational efficiency.

Yes, this context audit tool works without external dependencies, using internal scripts and references to analyze markdown documentation and conversation transcripts for anti-patterns and friction signals.