← Back to all free guides
A free guide by Mariah Brunner↗ INSTAGRAM↗ TIKTOK

The Context
Cleanup Checklist

Anthropic cut over 80% of Claude Code's instructions and it got no worse. Here is the exact checklist to trim your own AI setup so it thinks better, not harder.

SourceAnthropic: The New Rules of Context Engineering

You commented TRIM, so here it is. Anthropic just published new guidance for the Claude 5 generation, and buried in it was a wild admission: they deleted more than 80% of Claude Code's system prompt and saw no measurable drop in quality.

The lesson for the rest of us is simple. Most AI setups are not underfed, they are overloaded. Fewer, sharper instructions beat a wall of rules. This is the exact pass I run on my own setup in Claude, and you can do it in about ten minutes.

Run This PassThe Context Cleanup Checklist
  1. Delete every rule that states the obvious. "Write good code," "be helpful," "use correct grammar." Modern models already do this. Every line you cut is one less thing competing for attention.
  2. Turn rules into outcomes. Instead of ten style rules, say the result you want: "match the style of the file you are editing." That is Anthropic's own example. Outcomes travel further than rules.
  3. Kill the conflicts. Two instructions that fight, like "always be concise" and "explain every step," make AI hesitate and guess. Pick one.
  4. Break the one giant file into small ones. Do not dump everything into a single mega-prompt. Keep the always-true stuff short, and load the rest only when the task needs it. Anthropic calls this progressive disclosure.
  5. Show, do not describe. Replace a paragraph describing what you want with an actual example: a mockup, a sample output, a real file to copy. One reference beats three paragraphs.
  6. Move "how to do X" into the thing that does X. Setup steps for a specific task belong in that task's skill or project, not in your top-level instructions.
  7. Keep only the gotchas. Your custom instructions should hold the surprising, get-it-wrong-otherwise stuff. Not a manual.
Copy ThisThe 5-Minute Cleanup Prompt

Do not want to audit by hand? Paste your current instructions (your custom instructions, your project prompt, or your CLAUDE.md) into Claude with this, and it hands you a trimmed version plus a list of what it cut.

The Context Cleanup Prompt
You are my context editor. I am going to paste my current AI instructions below. Your job is to make them shorter and sharper without losing anything that actually changes how you behave. Apply these rules: 1. Delete anything that states the obvious or that a capable model already does by default. 2. Rewrite rigid rules as the OUTCOME I actually want (for example, "match the style of the file you are editing" instead of ten formatting rules). 3. Find any instructions that conflict with each other and tell me how you would resolve them. 4. Separate the always-true instructions (keep these short) from task-specific ones I should only load when they are relevant. 5. Point out where a real example, mockup, or reference file would replace a long description. 6. Keep only the non-obvious gotchas I would get wrong if you were not told. Return three things: A) The trimmed instructions, ready to paste back in. B) A short list of what you cut and why. C) Anything you were unsure about, so I can make the call. Here are my current instructions: [PASTE YOUR CUSTOM INSTRUCTIONS, PROJECT PROMPT, OR CLAUDE.MD HERE]

The Real Win

You are not trying to write the perfect prompt anymore. You are curating context. Give AI less to fight through and clearer references, and it gets sharper on its own.

Want the bigger picture on this skill? I broke down the context engineer role here. And when a chat gets bloated mid-task, my Save Your Claude Context handoff resets it clean.