中文文档 | English
A curated collection of production-grade meta-prompts for building high-performance AI agents and structured reasoning systems across multiple domains.
THE META PROMPT is a specialized repository providing battle-tested, structured meta-prompt templates designed to transform general-purpose Large Language Models into domain-specific reasoning agents with enhanced reliability and precision. Unlike traditional prompt libraries, this collection focuses on meta-level instruction frameworks that enable systematic task decomposition, role-based reasoning, and self-improving agent architectures.
- Zero-Shot Specialization: Transform base LLMs into specialized agents without fine-tuning or extensive example sets
- Systematic Reasoning: Implement structured thought processes through hierarchical instruction frameworks
- Production-Ready: Validated templates suitable for deployment in professional workflows
- Domain Coverage: Comprehensive coverage across general-purpose agents, academic research, and literature analysis
- Model Agnostic: Compatible with GPT-4, Claude, Gemini, Llama, and other instruction-following models
- Prompt Engineers building complex AI workflows
- AI Researchers developing agent frameworks
- Developers implementing LLM-based automation systems
- Academic professionals requiring structured reasoning tools
- Product teams integrating AI capabilities into production environments
- Modular prompt structure with clear role definitions and capability boundaries
- Explicit reasoning step requirements for transparent decision-making
- Structured output formats using XML/JSON tags for downstream parsing
- Error handling and edge case recovery mechanisms
- Multi-turn conversation state management protocols
- Production-validated across multiple LLM providers
- Consistent performance verified through systematic testing
- Comprehensive inline documentation within each template
- Version-controlled with semantic release tracking
- Clear modification guidelines for customization
- Language localization support with maintained structural integrity
- Domain-specific variant generation methodology
- Community contribution pipeline with quality standards
- Text editor or Markdown viewer
- LLM interface with system message support (API or chat interface)
- OpenAI API (GPT-3.5, GPT-4, GPT-4 Turbo)
- Anthropic Claude (Claude 3 Opus/Sonnet/Haiku)
- Google Gemini (Pro, Ultra)
- Open-source models (Llama 2/3, Mistral, Qwen)
- Any LLM supporting structured system prompts
None. Repository contains standalone Markdown files with no runtime dependencies.
# Clone repository
git clone https://github.com/HenryChiao/THE_META_PROMPT.git
cd THE_META_PROMPT
# View available templates
ls -la *.md# Example: OpenAI API Integration
import openai
# Load meta-prompt template
with open('THE_META_AGENT_PROMPT.md', 'r', encoding='utf-8') as f:
system_prompt = f.read()
# Initialize agent
response = openai.ChatCompletion.create(
model="gpt-4",
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": "Your task-specific query"}
]
)# Example: Anthropic Claude Integration
import anthropic
with open('THE_ACADEMIC.md', 'r', encoding='utf-8') as f:
system_prompt = f.read()
client = anthropic.Anthropic(api_key="your-api-key")
message = client.messages.create(
model="claude-3-opus-20240229",
max_tokens=4096,
system=system_prompt,
messages=[
{"role": "user", "content": "Analyze the methodology of [research paper]"}
]
)Purpose: Universal reasoning framework for general-purpose agent construction
Core Capabilities:
- Multi-step task decomposition with explicit reasoning chains
- Tool use orchestration and result integration
- Error detection and recovery strategies
- Context-aware response generation
- Self-consistency verification mechanisms
Use Cases:
- Complex workflow automation
- Multi-tool coordination systems
- Conversational AI with reasoning transparency
- Decision support applications
Structural Components:
├── Role Definition (Agent identity and boundaries)
├── Capability Matrix (Supported operations and limitations)
├── Reasoning Protocol (Step-by-step thought process structure)
├── Tool Integration Layer (External system interaction patterns)
├── Output Specification (Response format requirements)
└── Error Handling (Failure modes and recovery procedures)
Purpose: Specialized framework for academic analysis and research workflows
Core Capabilities:
- Citation handling and source evaluation
- Research methodology compliance
- Hypothesis generation and testing frameworks
- Literature review synthesis
- Academic writing standards enforcement
Use Cases:
- Systematic literature reviews
- Research proposal development
- Methodology critique and improvement
- Academic paper analysis
- Evidence-based argumentation
Structural Components:
├── Research Role Definition (Academic agent identity)
├── Source Evaluation Protocols (Citation quality assessment)
├── Methodology Framework (Research process structure)
├── Evidence Synthesis (Cross-source integration)
├── Academic Output Standards (Scholarly writing requirements)
└── Ethical Guidelines (Research integrity constraints)
Purpose: Chinese language optimization with cultural context adaptation
Core Capabilities:
- Native Chinese reasoning patterns
- Cultural context awareness
- Language-specific idiom handling
- Regional regulatory compliance
- Cross-cultural communication protocols
Use Cases:
- Chinese market applications
- Multilingual workflow systems
- Cultural adaptation requirements
- Regional compliance automation
Structural Components:
├── 角色定义 (Role Definition in Chinese)
├── 能力边界 (Capability Boundaries)
├── 推理流程 (Reasoning Workflow)
├── 文化适配 (Cultural Adaptation Layer)
├── 输出规范 (Output Specifications)
└── 异常处理 (Error Handling)
Purpose: Specialized framework for literary text analysis and critique
Core Capabilities:
- Narrative structure analysis
- Thematic element extraction
- Stylistic feature identification
- Character development tracking
- Literary device recognition
Use Cases:
- Literary criticism automation
- Text analysis workflows
- Creative writing feedback
- Comparative literature studies
Purpose: Foundational template for custom agent development
Core Capabilities:
- Baseline reasoning structure
- Generic task handling patterns
- Extensible framework foundation
- Minimal constraint baseline
Use Cases:
- Custom agent prototyping
- Domain-specific adaptation base
- Educational prompt engineering reference
import anthropic
# Load academic agent template
with open('THE_ACADEMIC.md', 'r') as f:
academic_prompt = f.read()
client = anthropic.Anthropic()
# Task: Systematic review of AI safety papers
response = client.messages.create(
model="claude-3-opus-20240229",
max_tokens=8192,
system=academic_prompt,
messages=[{
"role": "user",
"content": """
Conduct a systematic review of AI alignment research
published in 2023-2024. Focus on:
1. Key methodological approaches
2. Empirical validation strategies
3. Open research questions
4. Citation network analysis
"""
}]
)import openai
with open('THE_META_AGENT_PROMPT.md', 'r') as f:
agent_prompt = f.read()
# Agent with calculator and web search tools
response = openai.ChatCompletion.create(
model="gpt-4-turbo",
messages=[
{"role": "system", "content": agent_prompt},
{"role": "user", "content": "Calculate ROI for Q4 2024 and compare with industry benchmarks"}
],
tools=[
{"type": "function", "function": {"name": "calculator"}},
{"type": "function", "function": {"name": "web_search"}}
]
)# Extend general template for legal document analysis
with open('General_meta_prompt.md', 'r') as f:
base_prompt = f.read()
# Add domain-specific constraints
legal_prompt = base_prompt + """
## Legal Document Analysis Extensions
### Domain Constraints
- Apply strict citation verification
- Maintain jurisdiction-specific compliance
- Flag potential legal ambiguities
- Preserve original document language
### Output Requirements
- Structured legal analysis format
- Referenced case law citations
- Risk assessment framework
- Compliance checklist
"""
# Deploy customized agent# Original
You are a general-purpose reasoning agent...
# Customized for Financial Analysis
You are a quantitative financial analyst specializing in
risk assessment and portfolio optimization...# Add domain-specific capabilities
## Financial Analysis Capabilities
- Time series forecasting
- Risk metric calculation (VaR, CVaR)
- Correlation analysis
- Regulatory compliance checking<!-- Original structured output -->
<response>
<analysis>...</analysis>
<conclusion>...</conclusion>
</response>
<!-- Extended for financial reporting -->
<financial_report>
<executive_summary>...</executive_summary>
<quantitative_analysis>...</quantitative_analysis>
<risk_assessment>...</risk_assessment>
<recommendations>...</recommendations>
<compliance_notes>...</compliance_notes>
</financial_report>Always validate customizations across multiple LLM providers:
providers = [
("openai", "gpt-4"),
("anthropic", "claude-3-opus-20240229"),
("google", "gemini-pro")
]
for provider, model in providers:
response = test_prompt(custom_prompt, provider, model)
validate_output_format(response)
measure_consistency(response)All templates follow this organizational pattern:
1. Role Definition
├── Primary Identity
├── Capability Scope
└── Operational Boundaries
2. Reasoning Framework
├── Task Analysis Protocol
├── Decomposition Strategy
├── Step-by-Step Execution
└── Verification Procedures
3. Tool Integration (if applicable)
├── Available Tools Catalog
├── Tool Selection Criteria
├── Result Interpretation
└── Error Handling
4. Output Specification
├── Format Requirements
├── Structured Data Schema
├── Quality Standards
└── Edge Case Handling
5. Meta-Instructions
├── Self-Improvement Triggers
├── Clarification Protocols
├── Escalation Procedures
└── Performance Monitoring
Templates include embedded usage notes:
<!-- Usage Note: This section defines core reasoning protocol -->
## Reasoning Framework
[... template content ...]
<!-- Example: For financial analysis tasks, emphasize quantitative validation -->Release Date: 2024-02
Core Templates:
- THE_META_AGENT_PROMPT.md (General agent framework)
- THE_ACADEMIC.md (Academic research framework)
- THE_META_AGENT_PROMPT_CN.md (Chinese localization)
- Literature_meta_prompt.md (Literary analysis framework)
- General_meta_prompt.md (Foundational template)
Validation Status:
- Tested across GPT-4, Claude 3, Gemini Pro
- Production deployment verified
- Community feedback integrated
Known Limitations:
- Template file size may require chunking for models with small context windows
- Some templates assume tool use capabilities not available in all LLM APIs
- Chinese template optimization focused on Simplified Chinese
Accepted Contributions:
- Prompt clarity improvements
- Domain-specific template variants
- Additional language localizations
- Bug fixes and edge case handling
- Performance optimization
- Documentation enhancements
Contribution Process:
# Fork repository
git clone https://github.com/YOUR_USERNAME/THE_META_PROMPT.git
cd THE_META_PROMPT
# Create feature branch
git checkout -b feature/domain-legal-agent
# Add your template or modifications
# Follow existing structure and naming conventions
# Test across multiple LLM providers
python test_suite.py --template your_template.md
# Submit pull request with validation resultsPull Request Template:
## Template Name
[e.g., Legal Document Analysis Agent]
## Purpose
[Brief description of use case]
## Validation Results
- [ ] Tested with GPT-4
- [ ] Tested with Claude 3
- [ ] Tested with Gemini Pro
- [ ] Output format validation passed
- [ ] Edge case handling verified
## Sample Output
[Attach representative output examples]
## Additional Notes
[Any specific considerations or limitations]All contributions must maintain:
- Clear hierarchical structure matching existing templates
- Explicit reasoning step requirements
- Structured output format definitions
- Comprehensive inline documentation
- No external dependencies
- Cross-provider compatibility
- Maintain professional, respectful communication
- Provide constructive feedback on submissions
- Credit original prompt engineering research appropriately
- Respect intellectual property and licensing terms
MIT License
Copyright (c) 2024 THE META PROMPT Contributors
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Project Maintainer: HenryChiao
Repository: https://github.com/HenryChiao/THE_META_PROMPT
Issue Tracker: https://github.com/HenryChiao/THE_META_PROMPT/issues
Contribution Discussions: https://github.com/HenryChiao/THE_META_PROMPT/discussions
For questions, feedback, or collaboration inquiries, please open an issue on GitHub or participate in community discussions.
This project builds upon research in meta-prompting frameworks and prompt engineering methodologies from the AI research community. Special recognition to contributors of structured prompting techniques and agent architecture patterns that informed these templates.
- Meta-Prompting: Enhancing Language Models with Task-Agnostic Scaffolding
- Prompt Engineering for Large Language Models: A Survey
- Constitutional AI: Harmlessness from AI Feedback