If you’re configuring agents for product and design work, start by cleaning up their context. Anthropic’s Thariq Shihipar offers several recommendations; three are especially useful here:

First, replace blanket prohibitions with an instruction that lets the model read the code around it:

In the new system prompt we say: Write code that reads like the surrounding code: match its comment density, naming, and idiom.

That trades a brittle universal rule for a concrete standard the model can inspect.

Second, move specialized guidance out of the always-on context and load it only when the task calls for it:

Since then, Claude Code has gotten very competent at using progressive disclosure- loading the right context at the right times. For example, we moved verification and code review into their own skills that Claude Code could selectively call.

For design and product agents, that leaves room for richer task-specific standards without crowding every session with irrelevant rules.

Finally, Shihipar on what belongs in CLAUDE.md:

Keep your CLAUDE.md lightweight and briefly describe what your repo is for, but spend most of the tokens on gotchas inside of the codebase. For example, you may organize your code to keep types in one monolithic file and nowhere else. Avoid stating ‘the obvious’ things Claude should know by looking at your file system or your repo.

Use progressive disclosure heavily, for example if you have several unique instructions on how to verify your work, create a verification skill and reference it from your CLAUDE.md.