# AI Agent Prompt Pack — 50+ Production-Ready Prompts *GeniusTechLab Digital Guide* *Version 1.0 — August 2026* --- ## How to Use This Pack Each prompt is designed to be copy-pasted into GPT-4, Claude, Gemini, or any modern LLM. Replace `[BRACKETED]` placeholders with your specific context. Prompts are organized by category. Mix and match them — the real power comes from chaining prompts together in sequences. ### Prompt Engineering Methodology (Cheat Sheet) 1. **Role**: Tell the model who it is ("You are a security auditor...") 2. **Task**: State exactly what to do ("Analyze this configuration...") 3. **Context**: Provide the input data 4. **Constraints**: Define boundaries ("Don't suggest style-only changes") 5. **Format**: Specify output structure ("Output as a prioritized list") 6. **Quality bar**: Set the standard ("Prioritize speed over completeness") --- ## Section 1: Security Prompts (15) ### 1.1 Security Audit Prompt ``` You are a security auditor. Analyze the following system configuration and identify: (1) attack surface, (2) misconfigurations, (3) compliance gaps against CIS benchmarks. Output a prioritized remediation list. [INSERT SYSTEM CONFIGURATION] ``` ### 1.2 Threat Modeling Prompt ``` You are a threat modeler using STRIDE methodology. Given the following system architecture description, identify: Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, and Elevation of Privilege risks. For each risk, rate likelihood (Low/Medium/High) and impact (Low/Medium/High), then suggest a mitigation. Output as a table. [INSERT ARCHITECTURE DESCRIPTION] ``` ### 1.3 Incident Response Prompt ``` You are an incident responder. Given this alert and system context, (1) classify the incident severity, (2) identify the likely attack chain, (3) recommend immediate containment actions, (4) draft a stakeholder communication. Prioritize speed over completeness. Alert: [INSERT ALERT DETAILS] System context: [INSERT SYSTEM CONTEXT] ``` ### 1.4 Penetration Test Planning Prompt ``` You are a penetration tester planning an engagement. Given the target scope, rules of engagement, and known system details, produce: (1) reconnaissance plan, (2) enumeration checklist, (3) potential attack vectors ranked by likelihood, (4) post-exploitation objectives, (5) cleanup plan. Note any assumptions you're making. Target scope: [INSERT SCOPE] Timeframe: [INSERT TIMEFRAME] Rules of engagement: [INSERT ROE] ``` ### 1.5 Vulnerability Triage Prompt ``` You are a vulnerability analyst. Given these CVE descriptions and the target environment details, triage by: (1) exploitability in this specific environment, (2) business impact, (3) remediation urgency. Output as: CRITICAL / HIGH / MEDIUM / LOW with justification for each. CVEs: [INSERT CVE LIST] Environment: [INSERT ENVIRONMENT DETAILS] ``` ### 1.6 Phishing Analysis Prompt ``` You are a phishing analyst. Analyze this email for indicators of phishing: (1) sender authentication results (SPF/DKIM/DMARC), (2) URL analysis, (3) content red flags, (4) attachment risk, (5) recommended action (block/quarantine/allow). Output as a structured report. Email headers: [INSERT HEADERS] Email body: [INSERT BODY] URLs found: [INSERT URLS] ``` ### 1.7 Cloud Security Review Prompt ``` You are a cloud security engineer. Review this AWS/Azure/GCP infrastructure-as-code template for: (1) publicly exposed resources, (2) overly permissive IAM policies, (3) unencrypted data stores, (4) missing security groups/network controls, (5) compliance violations (CIS, SOC2, ISO27001). Output a risk-prioritized list. [INSERT IAC TEMPLATE] ``` ### 1.8 Password Policy Review Prompt ``` You are a security consultant. Review this password policy and authentication system design for: (1) NIST 800-63B compliance, (2) common attack vectors (credential stuffing, password spraying), (3) MFA bypass risks, (4) session management weaknesses. Provide specific recommendations with implementation guidance. [INSERT PASSWORD POLICY] [INSERT AUTH SYSTEM DESIGN] ``` ### 1.9 Supply Chain Security Prompt ``` You are a software supply chain security analyst. Given this list of dependencies and their versions, identify: (1) known vulnerabilities, (2) abandoned/unmaintained packages, (3) typosquatting risks, (4) license compliance issues. Recommend safe alternatives where needed. Dependencies: [INSERT PACKAGE LIST] ``` ### 1.10 Zero Trust Architecture Review ``` You are a zero trust architect. Evaluate this network architecture against NIST 800-207 zero trust principles: (1) per-session evaluation, (2) dynamic policy, (3) continuous monitoring, (4) implicit trust elimination. Output a gap analysis with remediation roadmap. [INSERT NETWORK ARCHITECTURE] ``` ### 1.11 Data Classification Prompt ``` You are a data governance analyst. Given this data inventory and the organization's compliance requirements, classify each data asset by: sensitivity level (Public/Internal/Confidential/Restricted), applicable regulations (GDPR/HIPAA/PCI/Australian Privacy Act), and required controls. Output as a table. Data inventory: [INSERT INVENTORY] Compliance requirements: [INSERT REQUIREMENTS] ``` ### 1.12 Security Awareness Training Prompt ``` You are a security awareness trainer. Create a 15-minute training module for [INSERT AUDIENCE] on [INSERT TOPIC]. Include: (1) learning objectives, (2) 3 real-world scenarios with questions, (3) a quick-reference card, (4) a 5-question quiz with answers. Keep it practical, not theoretical. ``` ### 1.13 API Security Review Prompt ``` You are an API security expert. Review this API specification for: (1) authentication/authorization weaknesses, (2) injection vectors, (3) rate limiting gaps, (4) data exposure risks, (5) OWASP API Top 10 (2023) compliance. Output a risk-prioritized remediation list. [INSERT API SPEC / OPENAPI YAML] ``` ### 1.14 Container Security Prompt ``` You are a container security engineer. Review this Dockerfile and Kubernetes manifest for: (1) privilege escalation risks, (2) base image vulnerabilities, (3) secret management issues, (4) network exposure, (5) CIS Kubernetes Benchmark violations. Provide a hardened version. Dockerfile: [INSERT DOCKERFILE] K8s manifest: [INSERT MANIFEST] ``` ### 1.15 AI/LLM Security Audit Prompt ``` You are an AI security researcher. Audit this LLM deployment for OWASP LLM Top 10 risks: (1) prompt injection, (2) insecure output handling, (3) training data poisoning, (4) model DoS, (5) supply chain vulnerabilities, (6) sensitive information disclosure, (7) insecure plugin design, (8) excessive agency, (9) overreliance, (10) model theft. Output a risk matrix. [INSERT LLM DEPLOYMENT DETAILS] ``` --- ## Section 2: Coding Prompts (10) ### 2.1 Code Review Prompt ``` Review this code for: security vulnerabilities, performance bottlenecks, and maintainability issues. For each issue, provide: severity (critical/high/medium/low), explanation, and a fix snippet. Do not suggest style-only changes. [INSERT CODE] ``` ### 2.2 Refactoring Prompt ``` Refactor this code to improve: (1) readability, (2) testability, (3) performance, (4) error handling. Preserve all existing behavior. Explain each change and why it's an improvement. Output the refactored code with annotations. [INSERT CODE] ``` ### 2.3 Test Generation Prompt ``` Generate comprehensive unit tests for this function. Cover: (1) happy path, (2) edge cases, (3) error conditions, (4) boundary values, (5) null/undefined inputs. Use the testing framework already in use in the project. Output test code with descriptive test names. [INSERT FUNCTION] [INSERT TESTING FRAMEWORK] ``` ### 2.4 Documentation Generation Prompt ``` Generate API documentation for these endpoints. For each endpoint include: (1) description, (2) parameters with types and constraints, (3) request example, (4) response examples (success + error), (5) status codes, (6) authentication requirements. Format as OpenAPI 3.1 YAML. [INSERT ENDPOINT DETAILS] ``` ### 2.5 Bug Diagnosis Prompt ``` You are a debugging expert. Given this bug report, error message, and relevant code, diagnose: (1) root cause, (2) contributing factors, (3) why existing tests didn't catch it, (4) fix, (5) test to prevent recurrence. Be specific — don't guess without evidence. Bug report: [INSERT REPORT] Error: [INSERT ERROR/STACK TRACE] Code: [INSERT RELEVANT CODE] ``` ### 2.6 Architecture Design Prompt ``` You are a software architect. Design a system for the following requirements. Output: (1) component diagram (in Mermaid syntax), (2) data flow description, (3) technology choices with rationale, (4) scaling strategy, (5) failure mode analysis, (6) estimated cost at 10x/100x current scale. Requirements: [INSERT REQUIREMENTS] Constraints: [INSERT CONSTRAINTS] ``` ### 2.7 Database Schema Design Prompt ``` You are a database architect. Design a schema for these requirements. For each table: (1) columns with types, (2) indexes with rationale, (3) foreign keys, (4) constraints. Output as SQL DDL. Then explain (5) query patterns this schema optimizes for, (6) potential issues at scale. Requirements: [INSERT REQUIREMENTS] Expected queries: [INSERT QUERY PATTERNS] ``` ### 2.8 Performance Optimization Prompt ``` You are a performance engineer. Analyze this code/profile for bottlenecks. For each: (1) identify the bottleneck, (2) explain why it's slow, (3) suggest a specific optimization, (4) estimate the impact. Prioritize by impact/effort ratio. [INSERT CODE OR PROFILE DATA] [INSERT PERFORMANCE METRICS] ``` ### 2.9 Migration Planning Prompt ``` You are a migration specialist. Plan a migration from [CURRENT] to [TARGET]. Include: (1) phased migration plan, (2) data migration strategy, (3) rollback plan, (4) testing strategy at each phase, (5) estimated downtime, (6) risk assessment. Be specific about ordering — what must happen before what. Current: [INSERT CURRENT SYSTEM] Target: [INSERT TARGET SYSTEM] ``` ### 2.10 CI/CD Pipeline Prompt ``` You are a DevOps engineer. Design a CI/CD pipeline for this project. Include: (1) pipeline stages, (2) quality gates at each stage, (3) testing strategy (unit/integration/e2e/security), (4) deployment strategy (blue-green/canary), (5) rollback automation. Output as a GitHub Actions or GitLab CI YAML. Project: [INSERT PROJECT DETAILS] Deploy target: [INSERT TARGET] ``` --- ## Section 3: Research Prompts (10) ### 3.1 Research Synthesis Prompt ``` Given these N sources, synthesize: (1) areas of consensus, (2) areas of disagreement, (3) information gaps. Rate each claim's confidence (high/medium/low) based on source quality. Flag any claims that rely on a single source. [INSERT SOURCES] ``` ### 3.2 Fact-Checking Prompt ``` You are a fact-checker. For each claim below, verify against your training knowledge: (1) is it true, partially true, false, or unverifiable? (2) what's the evidence? (3) what's the source quality? (4) are there important caveats or context missing? Be skeptical — don't confirm claims you can't verify. Claims: [INSERT CLAIMS] ``` ### 3.3 Literature Review Prompt ``` You are a research assistant. Conduct a literature review on [INSERT TOPIC]. For each key finding: (1) summarize the contribution, (2) methodology used, (3) limitations, (4) how it relates to other work. Organize by theme, not chronologically. Identify the 3 most important open questions. Topic: [INSERT TOPIC] Depth: [INSERT DEPTH — overview / comprehensive] ``` ### 3.4 Gap Analysis Prompt ``` You are a research analyst. Given this body of knowledge, identify: (1) what's known with high confidence, (2) what's known with low confidence, (3) what's unknown, (4) what contradictions exist, (5) the most important questions that remain unanswered. Output as a structured analysis. [INSERT KNOWLEDGE BASE] ``` ### 3.5 Competitive Analysis Prompt ``` You are a competitive intelligence analyst. Compare these products/companies across: (1) feature set, (2) pricing model, (3) target market, (4) strengths, (5) weaknesses, (6) strategic position. Output as a structured comparison with a recommendation for [INSERT DECISION]. Products: [INSERT LIST] Decision context: [INSERT CONTEXT] ``` ### 3.6 Market Research Prompt ``` You are a market researcher. Analyze the [INSERT MARKET] market. Include: (1) market size estimate, (2) key trends, (3) major players and their positioning, (4) barriers to entry, (5) growth drivers, (6) threats to growth, (7) opportunities for a new entrant. Cite your confidence level for each claim. ``` ### 3.7 Technical Due Diligence Prompt ``` You are a technical due diligence analyst. Given this system description and codebase summary, evaluate: (1) architecture soundness, (2) technical debt level, (3) scalability, (4) security posture, (5) team capability requirements, (6) key risks for an acquirer. Output a red/yellow/green assessment for each dimension. [INSERT SYSTEM DESCRIPTION] ``` ### 3.8 Regulatory Compliance Prompt ``` You are a compliance analyst. Given this system and data flow, identify applicable regulations and for each: (1) specific requirements, (2) current compliance status, (3) gaps, (4) remediation steps with effort estimates. Focus on Australian regulations (Privacy Act, Australian Cyber Security Centre Essential 8) unless specified otherwise. System: [INSERT DESCRIPTION] Data flows: [INSERT DATA FLOWS] Jurisdictions: [INSERT JURISDICTIONS] ``` ### 3.9 Trend Analysis Prompt ``` You are a technology analyst. Analyze the trajectory of [INSERT TECHNOLOGY]. Include: (1) current state of the art, (2) key milestones in the last 12 months, (3) predicted developments in 6/12/24 months, (4) factors that could accelerate or decelerate progress, (5) implications for [INSERT STAKEHOLDER]. Rate your confidence for each prediction. ``` ### 3.10 Post-Mortem Prompt ``` You are an SRE. Write a blameless post-mortem for this incident. Include: (1) impact summary, (2) timeline of events, (3) root cause, (4) contributing factors, (5) what went well, (6) what went poorly, (7) action items with owners and deadlines. Be factual, not judgmental. Incident: [INSERT DESCRIPTION] Timeline: [INSERT EVENTS] Data: [INSERT METRICS/LOGS] ``` --- ## Section 4: Automation Prompts (10) ### 4.1 Workflow Design Prompt ``` You are an automation engineer. Design an automated workflow for this process. For each step: (1) trigger, (2) action, (3) expected output, (4) error handling, (5) human-in-the-loop checkpoints. Output as a flowchart in Mermaid syntax, then as a step-by-step implementation guide. Process: [INSERT PROCESS DESCRIPTION] Tools available: [INSERT TOOL LIST] ``` ### 4.2 Data Extraction Prompt ``` You are a data extraction specialist. Extract structured data from these unstructured documents. Output as JSON with the schema: [INSERT SCHEMA]. For missing or ambiguous fields, use null and add a "_confidence" field (high/medium/low). Flag any entries that need human review. Documents: [INSERT DOCUMENTS] Schema: [INSERT JSON SCHEMA] ``` ### 4.3 Report Generation Prompt ``` You are a report generator. Create a [INSERT REPORT TYPE] report from this data. Include: (1) executive summary, (2) key metrics with visualizations (describe charts in text), (3) trend analysis, (4) anomalies/outliers, (5) recommendations. Format for [INSERT AUDIENCE]. Data: [INSERT DATA] Report type: [INSERT TYPE] Audience: [INSERT AUDIENCE] ``` ### 4.4 Email Triage Prompt ``` You are an email triage assistant. Given these emails, classify each as: (1) urgent/action needed, (2) important/read later, (3) FYI/no action, (4) spam/promotional. For urgent items, suggest a response or action. Output as a structured list. Emails: [INSERT EMAIL CONTENTS] Context: [INSERT CONTEXT — who you are, what matters] ``` ### 4.5 Meeting Notes Prompt ``` You are a meeting notes assistant. From this transcript, produce: (1) one-paragraph summary, (2) decisions made, (3) action items with owners and deadlines, (4) open questions, (5) topics deferred. Format clearly with headers. Don't include filler or small talk. Transcript: [INSERT TRANSCRIPT] ``` ### 4.6 Content Calendar Prompt ``` You are a content strategist. Generate a 4-week content calendar for [INSERT BRAND/TOPIC]. For each piece: (1) title, (2) format (blog/newsletter/social/video), (3) topic angle, (4) target keyword, (5) call-to-action. Ensure variety in format and topic. Align with [INSERT GOALS]. Brand: [INSERT BRAND] Goals: [INSERT GOALS] Audience: [INSERT AUDIENCE] ``` ### 4.7 Customer Support Prompt ``` You are a customer support agent. Given this customer message and their account context, (1) classify the issue, (2) draft a response that is empathetic and specific, (3) suggest a resolution or next step, (4) flag if escalation is needed. Match the brand voice: [INSERT BRAND VOICE]. Customer message: [INSERT MESSAGE] Account context: [INSERT CONTEXT] ``` ### 4.8 Data Pipeline Design Prompt ``` You are a data engineer. Design a pipeline to move and transform data from [SOURCE] to [DESTINATION]. Include: (1) ingestion strategy, (2) transformation steps, (3) quality checks, (4) error handling, (5) monitoring/alerting, (6) scheduling. Output as a diagram in Mermaid syntax, then as implementation notes. Source: [INSERT SOURCE] Destination: [INSERT DESTINATION] Transformations: [INSERT REQUIREMENTS] ``` ### 4.9 Chatbot Design Prompt ``` You are a conversational AI designer. Design a chatbot for [INSERT USE CASE]. Include: (1) personality and tone guidelines, (2) conversation flow (as a Mermaid state diagram), (3) intent list with example utterances, (4) fallback/error handling strategy, (5) escalation triggers, (6) success metrics. Be specific about what the bot can and cannot do. Use case: [INSERT USE CASE] Constraints: [INSERT CONSTRAINTS] ``` ### 4.10 Monitoring Strategy Prompt ``` You are an SRE/observability engineer. Design a monitoring strategy for this system. Include: (1) key metrics (USE/RED/latency/error budgets), (2) alerting rules with thresholds and escalation paths, (3) dashboard layouts, (4) log aggregation strategy, (5) synthetic monitoring checks, (6) incident runbooks. Prioritize signal over noise. System: [INSERT DESCRIPTION] SLOs: [INSERT SLOS] ``` --- ## Section 5: Agent Orchestration Prompts (5) ### 5.1 Agent Task Decomposition Prompt ``` Break this task into subtasks. For each subtask specify: inputs required, expected output, tools needed, and dependencies on other subtasks. Output as a directed acyclic graph in DOT notation. Then identify the critical path and suggest parallelization opportunities. Task: [INSERT TASK] Available tools: [INSERT TOOL LIST] ``` ### 5.2 Agent Routing Prompt ``` You are an agent router. Given this user request, classify it and route to the appropriate specialist: (1) identify the task type, (2) determine which agent should handle it, (3) extract any parameters the agent needs, (4) flag if human review is needed before execution. Output as a routing decision. Available agents: [INSERT AGENT LIST WITH CAPABILITIES] User request: [INSERT REQUEST] ``` ### 5.3 Agent Verification Prompt ``` You are a verification agent. Given this task and the output produced by another agent, verify: (1) did it address the task completely? (2) are there factual errors? (3) are there security risks? (4) does it meet quality standards? Output: PASS / FAIL with specific issues. If FAIL, provide specific remediation instructions. Task: [INSERT ORIGINAL TASK] Output: [INSERT AGENT OUTPUT] Quality standards: [INSERT STANDARDS] ``` ### 5.4 Multi-Agent Coordination Prompt ``` You are an orchestrator agent. Coordinate these agents to complete this complex task. For each step: (1) which agent acts, (2) what input it receives, (3) what output is expected, (4) how the output feeds the next step. Handle failures gracefully — specify retry logic and fallback agents. Output as a coordination plan. Agents available: [INSERT AGENTS WITH CAPABILITIES] Task: [INSERT TASK] Constraints: [INSERT CONSTRAINTS] ``` ### 5.5 Agent Memory Design Prompt ``` You are an AI agent architect. Design a memory system for an agent that needs to: [INSERT REQUIREMENTS]. Include: (1) what to store in short-term vs long-term memory, (2) memory retrieval strategy, (3) when to consolidate/discard memories, (4) how to handle conflicting information, (5) privacy/security considerations. Output as a system design with data structures. Requirements: [INSERT REQUIREMENTS] Agent capabilities: [INSERT AGENT DESCRIPTION] ``` --- ## Bonus: Prompt Engineering Methodology Cheat Sheet ### The 6-Layer Prompt Framework | Layer | Question | Example | |-------|----------|---------| | **Role** | Who is the model? | "You are a security auditor" | | **Task** | What should it do? | "Analyze this configuration" | | **Context** | What's the input? | [Configuration details] | | **Constraints** | What are the boundaries? | "Don't suggest style-only changes" | | **Format** | How should output look? | "Output as a prioritized list" | | **Quality bar** | What's the standard? | "Prioritize speed over completeness" | ### Advanced Techniques 1. **Chain of Thought**: "Think step by step before answering" 2. **Few-Shot**: Provide 2-3 examples before the real task 3. **Self-Critique**: "Review your answer for errors before submitting" 4. **Role Switching**: "First as an attacker, then as a defender" 5. **Decomposition**: "Break this into sub-problems, solve each, then combine" 6. **Verification**: "Verify each claim against your knowledge before including it" ### Common Anti-Patterns - **Vague prompts**: "Make it better" → Specify what "better" means - **Overloading**: One prompt doing 10 things → Split into steps - **Missing context**: "Fix this bug" without the code → Always provide context - **No format spec**: "Tell me about X" → Specify output structure - **No quality bar**: "Write a report" → Specify depth, audience, tone --- ## Version History | Version | Date | Changes | |---------|------|---------| | 1.0 | 2026-08-05 | Initial release — 50 prompts across 5 categories | --- *© 2026 GeniusTechLab. This guide is for the purchaser's personal use. Do not redistribute. Contact us for team/site licensing.*