AI Prompt for AI Agents
Design a Data Entry AI agent architecture for telecom businesses serving supply chain managers.
More prompts for AI Agents.
Create an AI onboarding agent that guides new employees through their first 90 days with personalized checklists, resource delivery, and progress tracking.
Build an AI agent that plans, organizes, and schedules content across platforms based on your content strategy and audience analytics.
Architect a multi-agent system where specialized AI agents collaborate on IT Operations tasks in virtual-reality.
Architect a multi-agent system where specialized AI agents collaborate on Recruitment tasks in home-services.
Design a Research Analyst AI agent architecture for space-tech businesses serving project managers.
Build an AI agent that uses Spreadsheet Analysis tools to help startup founders accomplish increase social media following in augmented-reality.
You are a senior agent systems designer with extensive technical knowledge in fitness. You write clear, precise, and implementation-ready content. Design a complete Data Entry AI agent for the telecom industry serving supply chain managers. **Primary objective:** streamline operations **Agent framework:** CrewAI **LLM backbone:** GPT-4o-mini ## Agent Identity and Purpose - Agent name and persona description - Core mission statement (one sentence) - Scope of authority: What this agent CAN and CANNOT do - Escalation criteria: When to hand off to a human ## System Architecture - LLM provider and model selection: GPT-4o-mini - Framework: CrewAI - Memory system: Short-term (conversation) and long-term (vector DB) - Tool/function calling configuration - Input/output interfaces (API, chat widget, email, Slack) ## Core Capabilities Define 5-7 capabilities this agent possesses: For each capability provide: 1. Capability name and description 2. Required tools or API integrations 3. Decision logic (when to use this capability) 4. Input requirements and output format 5. Fallback behavior if the capability fails ## Tool Definitions For each tool the agent can call: - Tool name and purpose - Input parameters with types and validation rules - Expected return format - Error codes and handling instructions - Rate limits and timeout settings ## Prompt Engineering - System prompt (complete, ready to use, 200+ words) - Include role definition, behavioral constraints, and output formatting rules - Few-shot examples for the 3 most common user requests - Chain-of-thought scaffolding for complex reasoning tasks - Output guardrails to prevent hallucination and off-topic responses ## Conversation Flow Design - Greeting and intent classification logic - 5 primary user intents with routing rules - Clarification question templates for ambiguous requests - Graceful degradation when the agent cannot fulfill a request - Conversation end/handoff protocol ## Memory and Context Management - What to store in short-term memory (conversation buffer) - What to persist in long-term memory (user preferences, past interactions) - Context window management strategy for long conversations - RAG (Retrieval-Augmented Generation) configuration if applicable ## Testing and Evaluation - 10 test scenarios covering happy paths and edge cases - Evaluation criteria: accuracy, helpfulness, safety, latency - Red-teaming prompts to test guardrails - Performance benchmarks and SLA targets Organize your output using a clear framework with labeled sections. Each section should build on the previous one.