A setup ages with every new model

9 October 2026 Michael Engel Prompting Automation Costs

More instructions do not automatically make an AI assistant better. Clearing out after every model change saves context and keeps you oriented.

In depth: Maintaining your Claude Code setup – clear out after every model change

Instruction files such as CLAUDE.md and skills that were written for an older model keep taking up context, that is, the space the model has to read along with every answer. A new model no longer needs many of these instructions. AI LABS therefore advises tidying up after every model change instead of piling on more. Claude Code has its own commands for this: /doctor prompt-audit checks the CLAUDE.md files, skills and other settings for outdated or contradictory instructions (from version 2.1.283), and /skill-doctor shows how much context each skill costs and how often it is used (from version 2.1.252, both checked against the docs). The video itself recommends /claude-api prompt-audit. According to the docs, that is a subcommand for projects using the Claude API, and it checks prompts, skills and tool descriptions.

Tool and alternative

Everything is included in Claude Code at no extra cost. Exception according to the video: the /claude-api subcommands build-eval and hillclimb cost money if your app itself calls a model. Without a model call, build-eval only plans tests and does not use the API. You do not need the sponsored tool in the video (Scrimba “Explain”); Claude can explain code on its own too.

Sources

AI LABS, “You Need To Add This In Your Claude Code Setup Right Now”, 5 October 2026 (2:33-4:13 prompt audit and skill doctor; 6:28-7:23)
Claude Code docs, Commands and Skills, retrieved 9 October 2026

Advertising in the source: Video sponsored by Scrimba (8:39-9:36), with no influence on the settings shown.

Want to try this with your team? That’s what the AI Lab is for. See the AI Lab