Unclear Why opusplan switched to the short prompt is unclear; it may be because it now resolves to a different underlying model.
The system prompt is the standing set of instructions Claude Code gives Claude at the start of every conversation. When you use the opusplan model setting, in the terminal or through the SDK, Claude now gets a short version of that prompt instead of the full one. Claude Code already had this short version; what changed is that opusplan now uses it.
The short prompt has:
- A "# Harness" section.
- One paragraph about risky actions.
- A condensed "# Memory" section.
These long sections of the full prompt are gone for opusplan: "Doing tasks", "Executing actions with care", "Tone and style", "Text output", and the long guide to the types of memory Claude can save.
The shorter prompt no longer carries an explicit list of destructive actions to be careful with, and its memory guidance is much briefer. If you use opusplan, Claude's behaviour may shift as a result.
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
* Fixed switching between Opus 5.5 and Sonnet 5.5 (`/model`, `opusplan`) rewriting earlier MCP tool announcements, which could drop earlier extended thinkingchangelog see the edit
Why `opusplan` switched to the short prompt is unclear; it may be because it now resolves to a different underlying model.
Anthropic's documentation has since written up opusplan, on Claude Code changelog.