agent-document-reviewer

Reviewing AGENTS.md files involves analyzing document structure, detecting bloat, and evaluating progressive disclosure to generate a prioritized improvement plan. This process computes quantitative metrics like line count and link ratios to identify quality gaps and suggest concrete reorganizations for better AI agent instructions.

Progressive disclosure in agent instruction documents is an optimization technique that structures information hierarchically to maintain token efficiency. It ensures AI agents receive necessary context without overwhelming them, evaluating link ratios and readability to determine if instructions should be split into separate referenced files.

Checking Claude.md instructions for redundancy requires quantitative analysis of line count, structure, and link ratios to detect content bloat. This evaluation identifies overlapping directives and provides concrete recommendations for reorganization, splitting guidance, and actionable improvements to optimize token efficiency.

Yes, you can audit agent instruction documents without external dependencies by using included scripts and references. The analysis operates independently on formats like AGENTS.md and Claude.md, computing metrics and evaluating reading impact to return a prioritized, actionable improvement plan directly.

The best way to measure reading impact of AI agent instructions is through quantitative metrics analysis, evaluating line count, structure, and link ratios. This scoring mechanism evaluates how effectively documents convey actionable guidance to AI agents, identifying quality gaps and suggesting concrete improvements.

You should split agent instruction documents when redundancy detection and progressive disclosure evaluation identify content bloat that hinders token efficiency. If quantitative analysis reveals excessive line counts or poor link ratios, the review provides concrete split guidance to reorganize instructions into separate referenced files.