Loading page content
Convert between AGENTS.md and CLAUDE.md, and lint either one for bloat: redundant rules, restatements of what the code already shows, and instructions that have never prevented a mistake.
Convert between CLAUDE.md and AGENTS.md (the cross-tool standard read by Codex, Copilot, Cursor, Gemini CLI, and 20+ others), and lint the result for bloat. Auto-generated agent config files are the most cited anti-pattern in the ecosystem; this linter flags the lines that cost tokens without changing behavior.
This project uses TypeScript and Next.js with the App Router.- IMPORTANT: Always write clean, maintainable code with meaningful variable names.- Always follow best practices and handle errors gracefully.The linter's premise: every always-loaded sentence is a per-session tax, and generic advice is pure waste. Full argument in how to write a CLAUDE.md and context engineering as FinOps. Audit the rest of your session with the context auditor.
Agent instruction files bloat by accretion. Every incident adds a rule, nothing is ever removed, and eighteen months later the file is 4,000 tokens loaded on every session, most of it defending against mistakes that happened once under conditions that no longer exist. The cost is real and recurring, and it is invisible because nobody re-reads the file.
The lint asks a single question of each rule: can you name the mistake it prevents? Rules that survive that question are worth their tokens. Rules restating the framework, describing the directory layout, or repeating something the code makes obvious are not. The model can read the repository.