🧹 Lancetta how-to #3 — stop paying for what your agents read before you type

🧹 Every session of Claude Code or Codex re-reads CLAUDE.md, AGENTS.md, what they import, the rules and the memory index before your first prompt. And pays for it again, every time.

Lancetta's Maintenance pane lists those files, shows what a start costs against the size where Claude Code warns, and runs two dozen checks:

🔍 an AGENTS.md that Claude Code never reads
🔗 an import that loads nothing
🪝 a hook that can't start
📏 a memory index so long its newest entries never load

Then the fixes:

✅ One right answer? You see the whole change first, and it's applied whole or not at all, one by one or with Fix All.
🧠 Needs judgement? Fix with Claude Code (it starts in plan mode) or Fix with Codex opens Terminal with a first message already written.

Nothing is written before you've seen it.

📖 The full walkthrough, with the routine I use: https://lancetta.substack.com

Free · universal binary · macOS 15+ · signed and notarized

👉 Try it / see what's new: https://lancetta.app

💬 Questions and ideas: https://discord.gg/rdWu5yFCR6

What's the strangest thing a check has found in your setup? 👀