Anthropic updated the Claude.ai system prompt with the release of Claude Opus 4.7 (April 16, 2026). Key changes include: renaming the developer platform to 'Claude Platform', adding Claude in PowerPoint as a new agent tool, a greatly expanded child safety section with a new XML tag, new guidance to make Claude less pushy and more action-oriented (prefer acting over asking clarifying questions), a new `tool_search` mechanism to check for deferred tools before claiming a capability is unavailable, instructions to keep responses more concise, new disordered eating safety guidelines, protection against yes/no screenshot attacks on contested topics, and removal of the Trump presidency clarification (now covered by the model's updated knowledge cutoff of January 2026). The post also reveals the full list of tool descriptions available in the Claude chat UI by asking Claude directly.
Sort: