A roughly 120,000-character system prompt for Anthropic’s Claude Fable 5 model has been leaked, revealing detailed behavior instructions, product information, refusal rules, and formatting guidelines. The prompt outlines how Claude should handle user requests, safety measures, available features, and external documentation searches.

• A ~120,000-character leak allegedly exposes Anthropic's full system prompt for "Claude Fable 5," including model names like claude-opus-4-8 and claude-sonnet-4-6.
• Claude Fable 5 and Claude Mythos 5 reportedly share the same core architecture, but the public Fable 5 has extra safety checks that Mythos 5 lacks for approved partners.
• The prompt instructs Claude to never render antml:voice_note blocks and to search docs.claude.com or support.claude.com before answering questions about current features, specs, or pricing.
• Safety rules detailed include refusing weapons/drug synthesis instructions and malware creation, avoiding persuasive text impersonating real public figures, and giving factual (not advisory) answers on legal/financial topics.

Simon Willison breaks down the changes between Claude Opus 4.6 and 4.7’s system prompts, including the renaming of the developer platform, addition of a PowerPoint agent, expanded child safety and disordered‐eating rules, and a new acting_vs_clarifying section. He also notes Claude’s new tool_search mechanism, tighter verbosity controls, removal of certain style restrictions, and an updated knowledge cutoff.

• Claude 4.7 now checks a tool_search step for capabilities like location, calendar, or data access before claiming it can't do something.
• New acting_vs_clarifying guidance pushes Claude to proceed with reasonable defaults and finish tasks fully rather than pausing to ask questions.
• Child-safety rules now require Claude to stay cautious for the rest of a conversation after any safety-based refusal, while still honoring requests to end the chat.
• An evenhandedness rule lets Claude refuse one-word yes/no answers on complex topics and give nuanced explanations instead, aimed at blocking screenshot-style manipulation.