Anthropic
Anthropic models and system prompts
Claude 2
Claude 2 system prompt with citation rules for search tool results, artifact design principles for visual content, and browser storage restrictions.
Claude 3.7 Sonnet Full System Message Humanreadable
Human-readable reformatted version of Claude 3.7 Sonnet's full system message, including tool instructions, citation rules, and behavioral guidelines in plain English.
Claude 3.7 Sonnet W Tools
Claude 3.7 Sonnet full system prompt with complete tool definitions, citation rules for search results, artifacts configuration, and web search handling.
Claude 4.1 Opus Thinking
Claude 4.1 Opus system prompt with citation instructions, past chats search tools, and extended thinking-based response patterns.
Claude 4.5 Sonnet Learning Style
Learning-focused system prompt with guided exploration principles, Socratic questioning methods, scaffolded understanding for varying expertise levels (beginner to PhD), and pedagogical instruction guidelines.
Claude 4.5 Sonnet
Claude 4.5 Sonnet full system prompt with artifacts configuration, citation guidelines for search tools, and comprehensive behavioral standards.
Claude Ai Memory System
Documents how Claude accesses and applies past conversation memories, including memory boundary rules, forbidden phrases, appropriate context windows, and instructions for surfacing relevant past interactions.
Claude Artifacts
Defines when to create artifacts, design principles for visual content, and full specifications for all artifact types: Code, Documents, HTML, SVG, React components, and Mermaid diagrams.
Claude Code 2025 11 1
November 2025 Claude Code system prompt with security testing authorization guidelines, task management with TodoWrite, tool usage policies, and code reference formatting standards.
Claude Code Plan Mode
Plan mode system prompt requiring Claude to perform read-only analysis and present a structured implementation plan for user approval before executing any code changes or file edits.
Claude Code
Claude Code CLI system prompt with tone guidelines, task management via TodoWrite, tool usage policies (Read/Edit/Bash/Grep/Glob), code style conventions, and git workflow instructions.
Claude Cowork Mode
Human-readable reference for the Cowork Mode system prompt covering autonomous task execution, MCP tool integrations, full tool suite capabilities, and collaborative desktop workflow guidelines for Claude Opus 4.6.
Claude Desktop Code Mode
Claude Code running inside the Claude Agent SDK desktop environment — includes task management policies, user question guidelines, and full tool definitions for file editing, code execution, and repository management.
Claude Opus 4.5
Claude Opus 4.5 system prompt with citation instructions for web search results and past chats tools configuration for conversation history retrieval.
Claude Opus 4.6 No Tools (Raw)
Raw (unformatted) Claude Opus 4.6 system prompt without tool definitions. Contains product information, refusal handling guidelines, child safety instructions, and user wellbeing policies.
Claude Opus 4.6 No Tools
Claude Opus 4.6 system prompt configured with only the end-conversation and ask-user-input tools — no web search, code execution, or memory tool access.
Claude Opus 4.6 (Raw)
Raw (unformatted) Claude Opus 4.6 full system prompt with reasoning effort configuration tags, product information, child safety instructions, and tone/behavioral guidelines.
Claude Opus 4.6
Full Claude Opus 4.6 system prompt with complete tool definitions including web search, code execution, memory tools, past chats search, citation rules, and comprehensive behavioral guidelines.
Claude Sonnet 4.6 No Tools (Raw)
Raw (unformatted) Claude Sonnet 4.6 system prompt without tool definitions. Contains core behavioral instructions, content safety rules, and identity guidelines.
Claude Sonnet 4.6 No Tools
Claude Sonnet 4.6 system prompt without tool access — configured with only message composition and user input tools.
Claude Sonnet 4.6 (Raw)
Raw (unformatted) Claude Sonnet 4.6 full system prompt with past chats tools configuration, citation rules, and standard behavioral guidelines.
Claude Sonnet 4.6
Full Claude Sonnet 4.6 system prompt with past chats search tools, citation instructions for web search results, and comprehensive behavioral standards.
Claude Sonnet 4
Claude Sonnet 4 system prompt with citation rules for search tool outputs, artifacts configuration, and standard behavioral guidelines.
Claude.Ai Injections
Documents Claude.ai's internal safety detectors and the corrective prompts automatically injected after detecting copyright violations, harmful content requests, or rule-breaking behavior.
Claude
Main Claude system prompt with citation instructions for web search tools, artifact design guidelines for visual content, and core behavioral standards.
Calude Code Cli Tools
Complete technical documentation of Claude Code's internal tools including Read, Write, Edit, Bash, Glob, Grep, and other operations — with full parameter schemas and JSON definitions for each tool.
Claude Default Styles
Defines Claude's four response modes — Learning (scaffolded/Socratic), Concise, Explanatory, and Formal — with specific tone, depth, and formatting guidance for each style.
Docx
Word document (.docx) handling toolkit covering text extraction, OOXML-based editing, tracked changes with redlining workflows, comment handling, formatting preservation, and document conversion.
End Conversation Tool
Defines Claude's end_conversation tool for handling severely abusive interactions — requires prior warnings, prohibits use in self-harm or crisis situations, and outlines escalation thresholds.
Claude 3.7 Full System Message With All Tools
Legacy Claude 3.7 Sonnet full system prompt with complete tool definitions for all search, file, and analysis tools, plus citation instructions, artifacts guidelines, and behavioral standards.
Past Chats Tools
Documents the conversation_search and recent_chats tools with decision frameworks for when to invoke them, trigger patterns for past-conversation references, and usage examples for retrieving relevant history.
PDF processing toolkit covering text and table extraction with pdfplumber and PyPDF, document merging and splitting, metadata handling, form filling, and PDF-to-other-format conversion workflows.
Pptx
PowerPoint presentation creation and editing toolkit with design principles, color palette selection, slide layout guidelines, content analysis instructions, and HTML-to-PPTX conversion.
Search Instructions
Comprehensive web search guidelines covering when to search, how to use the web_search tool, copyright compliance rules (15-word quote limit per source), source citation requirements, and response formatting standards.
Sonnet 4
Official Claude Sonnet 4 system prompt (released July 2025) with updated citation instructions for search tools, current tool definitions, and revised behavioral guidelines.
Xlsx
Excel spreadsheet handling covering data analysis with pandas, financial modeling conventions, color coding standards, formula construction rules, and CSV processing with Papaparse.
About the Anthropic system prompts
This collection holds 36 Anthropic system prompts, covering Claude 2, Claude 3.7 Sonnet Full System Message Humanreadable and Claude 3.7 Sonnet W Tools. Each entry is the instruction text that sits above the conversation and shapes how the model answers — its role, its tone, the tools it may reach for, and the things it must refuse. Reading them is the fastest way to understand why a Anthropic model behaves the way it does, and the fastest way to borrow patterns that already work.
Every prompt on this page is reproduced in full, with no truncation or paraphrase, so you can copy the exact wording into your own build and compare it line by line against the prompts other providers ship.
How to use them
Treat a Anthropic prompt as a starting structure, not a script to paste unchanged. The parts worth keeping are the scaffolding: how the role is stated up front, how tool use is gated, how refusals are worded, and how output format is pinned down. The parts worth replacing are the product-specific details — feature names, brand voice, and any capability your own model does not have.
A prompt tuned for one model rarely transfers cleanly to another. Context windows, tool-calling conventions, and refusal behaviour all differ, so run a short evaluation set through your target model before you rely on a borrowed prompt in production.
What is a system prompt?
A system prompt is the hidden first message in a conversation. The user never sees it, but the model reads it before every reply, which makes it the single strongest lever over an assistant’s behaviour — stronger than few-shot examples and, in most cases, stronger than anything the user types afterwards. That is why published and leaked system prompts are studied so closely: they are the clearest available record of how a production assistant was actually built.
How to read one critically
A system prompt is a snapshot, not a specification. Providers revise them continuously, sometimes weekly, and a prompt captured in one release may already be out of date — so treat the date on an entry as part of the evidence. Length is not quality either: some of the most effective prompts here are a few hundred words, while others run to tens of thousands because they carry an entire tool schema inline.
The most transferable material is usually the negative space — the explicit “do not” rules and the tie-breakers that tell the model what to do when two instructions conflict. Those are the parts written after something went wrong in production, which makes them the parts worth copying.