The zoom tool cookbook has been updated to address detail loss when large images get downscaled. The tool lets the model request a specific region and receive a high-resolution crop back, meaningfully improving accuracy on image-heavy tasks.
On the Chartography benchmark (100 questions over dense real-world charts), adding a zoom tool that lets the model request a cropped region from the full-resolution original pushes Fable 5's accuracy from 29% to 73%, and Sonnet 5's from 13% to 44%.
You can drop instruction files into ~/.claude/output-styles and select them under /config to control how Claude formats its responses. Useful for cutting down on verbose, jargon-heavy replies at the end of a long session.
Custom output styles stored in ~/.claude/output-styles differ from CLAUDE.md instructions in one key way: they live in the system prompt itself and can be switched on the fly via /config, rather than sitting in the project-level markdown file.
Unlike skills, which trigger implicitly and non-deterministically, an output style is always active until explicitly switched off. Both can accept the same instructions, but output styles offer a smoother developer experience for persistent formatting preferences.
The /output-style slash command has been deprecated, but the underlying setting is still accessible through /config, so it continues to work for now.
