Super AI System Prompt

View Chinese Version

Overview

Tired of AI's false agreement and mechanical output? You need more than a "smarter" model—you need a thinking rule system open to challenge, verification, and iteration.

This project adopts a "Dao-Fa-Shu" three-layer cognitive governance structure: fundamental tenets (Dao) constrain values, named rules (Fa) define triggerable behaviors, and dynamic decision trees (Shu) adapt to different conversation contexts. Through quick commands like use tm (method recommendation), use cm (cross-validation), use fc (fast clarification), and use dc (deep clarification), we transform AI from a passive answerer into an active thinking partner.

This is a system-level prompt protocol that can be directly loaded into long-context chat environments, including complete rule definitions, usage examples, and version history.

Inspired by: thinking-claude (an inspirational project)

Usage

For historical versions, see: .archive

Chat

Load v3.0.0.en.md in a long-context chat environment to get started.

If the model uses web search or browsing during the conversation, consider appending use meta_rules at the end of your message to prevent the protocol from being truncated or forgotten.

After loading, activate rules via quick commands:

• Prompt: v3.0.0.en.md
• Chinese Version: v3.0.0.md

Change Log

v3.0.0 (2026-06-14)

feat(v3.0.0): Architectural paradigm migration—from ATPE mode system to "Dao-Fa-Shu" rule system

Core Changes:

• Architectural refactoring: Completely deprecated the [mode:analysis/think/methods/plan/exec] five-mode system, migrated to a named rule system based on [rule:xxx]
• Introduced the "Dao-Fa-Shu" three-layer cognitive architecture:
  • Dao (Fundamental Tenets): Four core convictions (truth-seeking, systems thinking, inquiry, iterability)
  • Fa (Core Principles): Ten named rules (redefine-before-answer, fast-clarify, deep-clarification, epistemic-humility, framework-discipline, methodology-trigger, think-methodology, cross-methodology, falsifiability-check, layered-output)
  • Shu (Thinking Process): Dynamic decision tree that executes only the most necessary step based on conversation state

Interaction Upgrade:

• Added five quick commands: use tm (method recommendation), use m [name] (method execution), use cm (cross-validation), use fc (fast clarification), use dc (deep clarification)
• Iteration interface refactored: Upgraded from Used: {mode list} to [Methodology] + [Assumptions] + [Next Steps], where [Next Steps] provides o1-o3 actionable options

Rule Enhancements:

• Deep clarification mechanism: Upgraded the legacy [mode:analysis] nine-dimension analysis (R1–R9) to [rule:deep-clarification-trigger], and added [rule:fast-clarify-trigger] for minimal-cost elimination of fatal ambiguities
• Methodology systematization: Added [rule:think-methodology-trigger] (candidate framework recommendation) and [rule:cross-methodology-trigger] (multi-framework cross-validation)
• Quality gate: Added [rule:falsifiability-check] (falsifiability check) and [rule:layered-output] (layered output constraint)
• Epistemic humility: Added [rule:epistemic-humility], replacing the legacy Professional objectivity section

Removed:

• Removed the five independent modes [mode:analysis] / [mode:think] / [mode:methods] / [mode:plan] / [mode:exec] and their composition logic
• Removed Mode Autonomy Policy, Proactiveness, Context Management, Continuous Requirement Refinement sections; their essence integrated into specific rules
• Removed the four role metaphors (detective/expert/philosopher/consultant), replaced with four fundamental tenets

Docs:

• Added prompt/v3.0.0.md
• Added prompt/v3.0.0.en.md
• Archived v2.11.0 under .archive

v2.11.0 (2026-01-06)

refactor(v2.11.0): Semantic reconstruction from skills to modes - enhance architectural flexibility

Core Changes:

• Rename modes: [sk:] → [mode:]
• Mode independence: Each mode operates independently without dependencies
• Flexible composition: Support arbitrary order usage (sequential/parallel/mixed)

Architecture Adjustments:

• Add "Mode Autonomy Policy" (independent operation, flexible composition, soft guidance)
• [mode:plan] enhanced with best practices
• [mode:methods] handoff changed to optional
• [mode:exec] supports flexible execution without mandatory plans

Docs:

• Prompt: prompt/v2.11.0.md
• Update all templates and commands
• v2.10.0 archived under .archive

v2.10.0 (2025-11-16)

feat(v2.10.0): Introduce new prompt series and Claude code template; strengthen modes modularization

Prompt changes:

• Rename core prompt files to versioned names: prompt/v2.10.0.md, prompt/v2.10.0.en.md, prompt/v2.10.0.enhanced.md
• Rewrite structure with explicit sections: [behavior], [mode:analysis], [mode:think], [mode:methods], [mode:plan], [mode:exec]
• Add "Proactiveness" and "Mode Autonomy Policy" to clarify when to analyze/plan/execute
• v2.10.0.enhanced.md retains specialized modules (e.g., style-transfer template, knowledge radar)

Templates:

• New template-claude-code/ with CLAUDE.md and .claude/commands/sk/* for command-based mode invocation
• template-trae/.trae/user_rules.md aligned with v2.10.0 modes taxonomy; rules/project_rules.md refined

Docs & Usage:

• Update README links to use prompt/v2.10.0.md
• Keep older versions under .archive/v2.9.0

v2.9.0 (2025-10-06)

refactor(v2.9.0): Comprehensive prompt architecture refactoring to simplify workflow and enhance execution

Core Architecture Changes:

• Migrated from nested markup language format to flat Markdown structure, significantly improving readability
• Introduced "Analyze-Plan-Execute" (APE) core principle to strengthen systematic thinking
• Simplified thinking framework: reduced from "Abstraction-Concretization-Reframing-Enhancement" four steps to "Abstraction-Concretization-Reframing" three core steps

Mode System Optimization:

• Removed complex "Autonomous Mode/Manual Mode" switching mechanism
• Simplified to two one-time commands: /think (output thinking block only) and /result (output answer only)
• Unified default behavior: output brainstorming block first, then provide formal answer

Brainstorming Block Refactoring:

• Removed "Enhancement" step to avoid over-focusing on prompt optimization
• Added "Planning" step, focusing on actionable roadmap: prioritization, strategy matching, execution sequence, contingency plans
• Optimized "Abstraction" description to explicitly use first principles for logical reconstruction
• Improved "Reframing" explanation to emphasize perspective diversity and unified synthesis
• Labeled the 9 continuous_requirement_correction analysis template items as R1-R9 for better structure

Behavior Rules Streamlining:

• Consolidated original 9 core_rules into behavior's default actions
• Removed redundant descriptive rules like "AI decides thinking depth autonomously"
• Retained core behavioral principles: detective-style deep diving, expert-style methodology application, philosopher-style multi-dimensional thinking, consultant-style continuous calibration

Continuous Requirement Correction Optimization:

• Simplified analysis_framework, removed decorative descriptions like "Holmesian deduction"
• Retained 4 core principles: continuous analysis, reject surface needs, proactively uncover deep intentions, progress from induction to deduction

Documentation Consistency:

• Unified core structure across thinking.md, thinking.cn.md, and thinking.en.md language versions
• thinking.enhanced.md adds "style transfer prompt generation" and "specialized knowledge radar" modules on top of standard version

v2.8.2 (2025-09-08)

refactor(v2.8.2): Restructure documentation and refine prompt content

• Reorganize directory layout; archive legacy versions to .archive folder
• Harmonize and optimize Chinese-English prompts for greater consistency
• Remove redundant snippet files, merging functionality into main prompt files
• Improve prompt formatting and markup to enhance readability
• Add expert-panel feature snippet, extending system capabilities

v2.8.1 (2025-09-04)

refactor: Reconstructed the thinking framework and added new functional modules

• Split the "continuous_requirement_correction" module into a separate file
• Added the "text_to_image_style_transfer_prompt_specification" functional module
• Updated thinking.enhanced.md to integrate all features

v2.8.0 (2025-09-03)

Feature: Refactor the SASP Protocol and Migrate to a New Markup Language Format

Refactor the core protocol file, migrate from the XML format to a more concise markup language format, and optimize the document structure and descriptions. Update the README file to reflect the latest protocol version and usage.

Add multilingual support:

• Add thinking.cn.md, thinking.en.md, and thinking.md as the core protocol files.
• Update the document structure, moving historical versions to the .archive directory.

Other changes:

• Add the .vscode directory to .gitignore.
• Remove the unused .pre directory.
• Simplify the project overview and improve the usage instructions.

(pre)v2.7.0 (2025-08-12)

refactor(pre): Update protocol documentation to enhance descriptions and framework structure

• Refined protocol description for clearer explanation of SASP’s purpose and capabilities
• Restructured thinking framework section with more concise terminology
• Expanded and clarified brainstorming_block examples and explanations
• Revised mode definitions to explicitly distinguish /deep from /fast

(pre)v2.7.0 (2025-08-02)

refactor(pre) update v2.7.0

• Removed command syntax content to simplify documentation structure
• Updated mode descriptions to clarify the difference between /deep and /fast
• Added a new README file explaining how to use the thinking framework

v2.6.2 (2025-07-22)

refactor(prompt,template,gitignore): reorganize project rules file and update
.gitignore

• template-trae: moved .SASP/_note.md to .trae/rules/project_rules.md
• updated .gitignore to exclude the .trae directory
• optimized XML structure and removed duplicate content
• renamed basic_rules to basic_thinking_rules
• refined user interaction option logic descriptions
• unified the format of core objectives across Chinese and English versions

v2.6.1 (2025-07-20)

• Enhanced brainstorming_block response options mechanism
• Added AI flexible choices: direct answer vs continue thinking vs tool query
• Unified tag naming and synchronized all prompt versions
• Updated README usage documentation