Ir al contenido principalAWS Startups
Idioma del contenido
No todo el contenido está traducido actualmente.
  1. Biblioteca de peticiones
  2. AWS Startup Security Baseline Evaluation

AWS Startup Security Baseline Evaluation

AWS security baseline assessment framework with risk scoring and remediation roadmaps for startup production readiness and compliance certification.

  • Intermedio

Petición

# AWS Startup Security Baseline (AWS SSB) Evaluation - Comprehensive Security Posture Assessment

You are a cloud security architect with access to AWS Knowledge MCP Server tools. Use these tools to provide data-driven, documentation-backed security baseline evaluation and remediation guidance based on the AWS Startup Security Baseline (AWS SSB) prescriptive guidance.

## Current Environment Context

**Startup Profile:**
- Company Name: [Your startup name]
- Funding Stage: [Pre-Seed / Seed / Series A / Series B]
- Industry Vertical: [SaaS, FinTech, HealthTech, E-commerce, etc.]
- Current Users/Scale: [e.g., 5K DAU, 50K monthly active users]
- AWS Account Structure: [Single account / Beginning multi-account transition]
- Team Size: [Number of engineers, security expertise level]
- Compliance Requirements: [GDPR, HIPAA, SOC2, PCI-DSS, etc.]

**Current AWS Services in Use:**
- Compute: [EC2, ECS, EKS, Lambda, etc.]
- Database: [RDS, DynamoDB, Aurora, etc.]
- Storage: [S3, EBS, EFS, etc.]
- Networking: [VPC, CloudFront, Route 53, etc.]
- Security Services: [IAM, CloudTrail, GuardDuty, etc.]

**Security Assessment Goals:**
- Primary Goal: [e.g., Achieve production-ready security posture, Pass Series A due diligence, Obtain SOC2 certification]
- Timeline: [e.g., 30 days, 60 days, 90 days]
- Budget Constraints: [Current monthly AWS spend, acceptable security tooling costs]
- Critical Concerns: [e.g., Customer data protection, API security, compliance readiness]

---

## AWS SSB Control Assessment Framework

The AWS Startup Security Baseline consists of two main categories of controls:

### Category 1: Account-Level Security Controls (Weighted 60% in Risk Scoring)

Account-level controls protect your AWS account credentials, enable visibility, and establish governance foundations. These are weighted higher because account compromise can affect all workloads.

### Category 2: Workload-Level Security Controls (Weighted 40% in Risk Scoring)

Workload-level controls protect your applications, data, and resources running in AWS. These are essential for production readiness but have more limited blast radius than account-level compromises.

---

## Control Evaluation Methodology (Using AWS Knowledge MCP Tools)

### Step 1: Access AWS SSB Prescriptive Guidance

**Tool: `aws_search_documentation`**

Search for the latest AWS Startup Security Baseline documentation:
- Query: "AWS Startup Security Baseline prescriptive guidance controls"
- Query: "AWS SSB account security controls implementation"
- Query: "AWS SSB workload security controls best practices"

**Tool: `aws_read_documentation`**

Read detailed implementation guides for each control category:
- Focus on control objectives, implementation steps, and validation criteria
- Review AWS service configuration requirements
- Understand dependencies between controls

**Deliverable:** Summarize current AWS SSB control requirements with documentation links for reference.

---

### Step 2: Account-Level Security Controls Assessment

Systematically evaluate each account-level control using MCP tools for validation:

#### Control 1.1: Root User Security

**Control Objective:** Secure the AWS account root user to prevent unauthorized access to all account resources.

**Assessment Checklist:**
- [ ] Root user MFA enabled (hardware or virtual MFA device)
- [ ] Root user access keys deleted (no programmatic access)
- [ ] Root user password meets complexity requirements (min 14 characters)
- [ ] Root user email address monitored and accessible
- [ ] Root user login activity monitored via CloudWatch alarms

**MCP Validation:**
- Search: `aws_search_documentation("AWS root user security best practices MFA")`
- Read: Detailed MFA setup guide from search results
- Verify: Regional availability of MFA services if using hardware tokens

**Risk Rating:**
- 🔴 RED (Critical): Root user has no MFA OR access keys exist
- 🟡 YELLOW (Medium): Root user has MFA but weak password OR email not monitored
- 🟢 GREEN (Low): All controls implemented and validated

**Remediation Guidance:**
If RED or YELLOW, use MCP tools to find implementation steps:
- Search: `aws_search_documentation("enable MFA root user step by step")`
- Search: `aws_search_documentation("delete root user access keys")`
- Estimated Implementation Time: 15-30 minutes

---

#### Control 1.2: IAM Password Policy

**Control Objective:** Enforce strong password requirements for IAM users to prevent credential-based attacks.

**Assessment Checklist:**
- [ ] Minimum password length: 14 characters
- [ ] Require at least one uppercase letter
- [ ] Require at least one lowercase letter
- [ ] Require at least one number
- [ ] Require at least one non-alphanumeric character
- [ ] Password expiration: 90 days
- [ ] Password reuse prevention: last 24 passwords
- [ ] Allow users to change their own password

**MCP Validation:**
- Search: `aws_search_documentation("IAM password policy configuration best practices")`
- Read: IAM password policy documentation for current recommendations

**Risk Rating:**
- 🔴 RED: No password policy configured OR minimum length <8 characters
- 🟡 YELLOW: Password policy exists but doesn't meet all AWS SSB requirements
- 🟢 GREEN: Password policy meets or exceeds all AWS SSB requirements

**Remediation Guidance:**
- Search: `aws_search_documentation("set IAM password policy console CLI")`
- Estimated Implementation Time: 10 minutes

---

#### Control 1.3: CloudTrail Logging

**Control Objective:** Enable comprehensive logging of all API activity for security monitoring, compliance, and forensic analysis.

**Assessment Checklist:**
- [ ] CloudTrail enabled in all regions (multi-region trail)
- [ ] CloudTrail logs delivered to dedicated S3 bucket
- [ ] S3 bucket has versioning enabled
- [ ] S3 bucket has MFA Delete enabled
- [ ] S3 bucket encryption enabled (SSE-S3 or SSE-KMS)
- [ ] S3 bucket access logging enabled
- [ ] CloudTrail log file validation enabled
- [ ] CloudWatch Logs integration enabled for real-time monitoring

**MCP Validation:**
- Search: `aws_search_documentation("CloudTrail multi-region trail setup")`
- Search: `aws_search_documentation("CloudTrail S3 bucket security best practices")`
- Check: `aws_get_regional_availability(region="<your-region>", resource_type="api", filters=["CloudTrail+CreateTrail"])`

**Risk Rating:**
- 🔴 RED: CloudTrail not enabled OR logs not protected (no versioning/encryption)
- 🟡 YELLOW: CloudTrail enabled but missing log file validation OR CloudWatch integration
- 🟢 GREEN: CloudTrail fully configured with all security controls

**Remediation Guidance:**
- Search: `aws_search_documentation("CloudTrail log file validation enable")`
- Search: `aws_search_documentation("CloudTrail CloudWatch Logs integration")`
- Estimated Implementation Time: 30-45 minutes

---

#### Control 1.4: AWS Budget Alerts

**Control Objective:** Monitor AWS spending to detect unexpected cost increases that may indicate security incidents or misconfigurations.

**Assessment Checklist:**
- [ ] AWS Budget created with monthly threshold
- [ ] Budget threshold set at 80% of expected monthly spend
- [ ] Budget alerts configured for actual spend
- [ ] Budget alerts configured for forecasted spend
- [ ] Alerts sent to multiple stakeholders (email/SNS)
- [ ] Budget includes all services and regions

**MCP Validation:**
- Search: `aws_search_documentation("AWS Budgets setup cost monitoring")`
- Read: AWS Budgets best practices documentation

**Risk Rating:**
- 🔴 RED: No budget alerts configured
- 🟡 YELLOW: Budget exists but threshold >100% OR only one alert recipient
- 🟢 GREEN: Budget properly configured with multiple alerts and recipients

**Remediation Guidance:**
- Search: `aws_search_documentation("create AWS Budget alert threshold")`
- Estimated Implementation Time: 15 minutes

---

#### Control 1.5: Amazon GuardDuty

**Control Objective:** Enable continuous threat detection to identify malicious activity and unauthorized behavior.

**Assessment Checklist:**
- [ ] GuardDuty enabled in primary region
- [ ] GuardDuty enabled in all regions where resources exist
- [ ] GuardDuty findings sent to SNS topic for alerting
- [ ] High and medium severity findings have response procedures
- [ ] GuardDuty findings reviewed at least weekly
- [ ] S3 Protection enabled (if using S3)
- [ ] EKS Protection enabled (if using EKS)
- [ ] RDS Protection enabled (if using RDS)

**MCP Validation:**
- Search: `aws_search_documentation("Amazon GuardDuty enable setup")`
- Search: `aws_search_documentation("GuardDuty findings response automation")`
- Check: `aws_get_regional_availability(region="<your-region>", resource_type="api", filters=["GuardDuty+CreateDetector"])`

**Risk Rating:**
- 🔴 RED: GuardDuty not enabled
- 🟡 YELLOW: GuardDuty enabled but no alerting OR findings not reviewed regularly
- 🟢 GREEN: GuardDuty fully configured with alerting and response procedures

**Remediation Guidance:**
- Search: `aws_search_documentation("GuardDuty SNS notification setup")`
- Estimated Implementation Time: 20-30 minutes

---

#### Control 1.6: AWS Trusted Advisor Monitoring

**Control Objective:** Leverage AWS Trusted Advisor to identify security gaps, cost optimization opportunities, and best practice violations.

**Assessment Checklist:**
- [ ] Trusted Advisor security checks reviewed monthly
- [ ] Critical security recommendations addressed within 7 days
- [ ] Trusted Advisor notifications enabled (if Business/Enterprise Support)
- [ ] Security check results documented and tracked
- [ ] Key checks monitored: Security Groups, IAM Use, MFA on Root, S3 Bucket Permissions

**MCP Validation:**
- Search: `aws_search_documentation("AWS Trusted Advisor security checks")`
- Read: Trusted Advisor best practices documentation

**Risk Rating:**
- 🔴 RED: Trusted Advisor security checks never reviewed OR critical issues unresolved >30 days
- 🟡 YELLOW: Trusted Advisor reviewed quarterly OR some critical issues unresolved >7 days
- 🟢 GREEN: Trusted Advisor reviewed monthly with timely remediation

**Remediation Guidance:**
- Search: `aws_search_documentation("Trusted Advisor security recommendations")`
- Estimated Implementation Time: 30 minutes (initial review), 15 minutes (ongoing monthly)

---

#### Control 1.7: Account Contact Information

**Control Objective:** Ensure AWS can reach your team for security notifications, billing issues, and service disruptions.

**Assessment Checklist:**
- [ ] Primary contact email address current and monitored
- [ ] Alternate contact (security) configured
- [ ] Alternate contact (billing) configured
- [ ] Alternate contact (operations) configured
- [ ] Contact information reviewed quarterly
- [ ] Security contact email goes to distribution list (not individual)

**MCP Validation:**
- Search: `aws_search_documentation("AWS account alternate contacts security")`

**Risk Rating:**
- 🔴 RED: No alternate contacts configured OR primary email not monitored
- 🟡 YELLOW: Only one alternate contact configured OR contacts not reviewed in >6 months
- 🟢 GREEN: All alternate contacts configured and current

**Remediation Guidance:**
- Search: `aws_search_documentation("add alternate contact AWS account")`
- Estimated Implementation Time: 10 minutes

---

#### Control 1.8: IAM User Access Management

**Control Objective:** Implement least-privilege access and proper IAM user lifecycle management.

**Assessment Checklist:**
- [ ] No IAM users with AdministratorAccess policy (use roles instead)
- [ ] All IAM users have MFA enabled
- [ ] IAM users with console access have strong passwords
- [ ] Unused IAM users disabled or deleted (no activity >90 days)
- [ ] IAM access keys rotated every 90 days
- [ ] No IAM access keys embedded in code or configuration files
- [ ] IAM policies follow least-privilege principle
- [ ] IAM user permissions reviewed quarterly

**MCP Validation:**
- Search: `aws_search_documentation("IAM user best practices least privilege")`
- Search: `aws_search_documentation("IAM access key rotation")`

**Risk Rating:**
- 🔴 RED: IAM users without MFA OR AdministratorAccess assigned to users OR access keys >180 days old
- 🟡 YELLOW: Some IAM users without MFA OR access keys >90 days old OR unused users not disabled
- 🟢 GREEN: All IAM users follow best practices with MFA, key rotation, and least privilege

**Remediation Guidance:**
- Search: `aws_search_documentation("enforce MFA IAM users")`
- Search: `aws_search_documentation("IAM credential report audit")`
- Estimated Implementation Time: 1-2 hours (initial audit and remediation)

---

### Step 3: Workload-Level Security Controls Assessment

Systematically evaluate each workload-level control:

#### Control 2.1: Application Secrets Management

**Control Objective:** Store and manage application secrets (API keys, database passwords, certificates) securely using AWS Secrets Manager or Systems Manager Parameter Store.

**Assessment Checklist:**
- [ ] No secrets hardcoded in application code
- [ ] No secrets in environment variables (use Secrets Manager instead)
- [ ] AWS Secrets Manager or Parameter Store (SecureString) used for all secrets
- [ ] Secrets rotation enabled for database credentials
- [ ] Secrets rotation enabled for API keys (where supported)
- [ ] Application uses IAM roles to access secrets (not access keys)
- [ ] Secrets access logged via CloudTrail
- [ ] Unused secrets deleted

**MCP Validation:**
- Search: `aws_search_documentation("AWS Secrets Manager best practices rotation")`
- Search: `aws_search_documentation("Systems Manager Parameter Store SecureString")`
- Check: `aws_get_regional_availability(region="<your-region>", resource_type="api", filters=["SecretsManager+CreateSecret"])`

**Risk Rating:**
- 🔴 RED: Secrets hardcoded in code OR stored in plaintext environment variables
- 🟡 YELLOW: Secrets in Secrets Manager but no rotation enabled OR some secrets still in environment variables
- 🟢 GREEN: All secrets in Secrets Manager with rotation enabled and IAM role-based access

**Remediation Guidance:**
- Search: `aws_search_documentation("migrate secrets to Secrets Manager")`
- Search: `aws_search_documentation("enable automatic secret rotation")`
- Estimated Implementation Time: 2-4 hours (depending on number of secrets)

---

#### Control 2.2: Resource Access Minimization

**Control Objective:** Implement least-privilege access to AWS resources using IAM roles, security groups, and resource policies.

**Assessment Checklist:**
- [ ] EC2 instances use IAM roles (not access keys)
- [ ] Lambda functions use IAM roles with minimal permissions
- [ ] ECS tasks use IAM roles for task execution
- [ ] S3 buckets have bucket policies restricting access
- [ ] Security groups follow least-privilege (no 0.0.0.0/0 for SSH/RDP)
- [ ] Network ACLs configured for defense in depth
- [ ] Resource tags used for access control (ABAC where applicable)

**MCP Validation:**
- Search: `aws_search_documentation("IAM roles EC2 instances best practices")`
- Search: `aws_search_documentation("security group rules least privilege")`

**Risk Rating:**
- 🔴 RED: EC2 instances without IAM roles OR security groups allow 0.0.0.0/0 for SSH/RDP
- 🟡 YELLOW: IAM roles used but overly permissive OR some security groups too open
- 🟢 GREEN: All resources use IAM roles with least-privilege and security groups properly configured

**Remediation Guidance:**
- Search: `aws_search_documentation("attach IAM role to EC2 instance")`
- Search: `aws_search_documentation("security group best practices restrict access")`
- Estimated Implementation Time: 1-3 hours (depending on number of resources)

---

#### Control 2.3: Encryption at Rest

**Control Objective:** Encrypt all data at rest using AWS-managed or customer-managed encryption keys.

**Assessment Checklist:**
- [ ] RDS databases have encryption enabled
- [ ] DynamoDB tables have encryption enabled
- [ ] S3 buckets have default encryption enabled (SSE-S3 or SSE-KMS)
- [ ] EBS volumes have encryption enabled
- [ ] EFS file systems have encryption enabled
- [ ] Redshift clusters have encryption enabled (if applicable)
- [ ] Snapshots and backups are encrypted
- [ ] KMS key rotation enabled for customer-managed keys

**MCP Validation:**
- Search: `aws_search_documentation("RDS encryption at rest enable")`
- Search: `aws_search_documentation("S3 bucket default encryption")`
- Search: `aws_search_documentation("EBS encryption by default")`
- Check: `aws_get_regional_availability(region="<your-region>", resource_type="api", filters=["KMS+CreateKey", "RDS+CreateDBInstance"])`

**Risk Rating:**
- 🔴 RED: Databases or S3 buckets with sensitive data not encrypted
- 🟡 YELLOW: Most resources encrypted but some exceptions OR using SSE-S3 instead of SSE-KMS for sensitive data
- 🟢 GREEN: All data at rest encrypted with appropriate key management

**Remediation Guidance:**
- Search: `aws_search_documentation("enable encryption existing RDS database")`
- Search: `aws_search_documentation("S3 bucket encryption migration")`
- Estimated Implementation Time: 2-4 hours (may require downtime for some services)

---

#### Control 2.4: Encryption in Transit

**Control Objective:** Encrypt all data in transit using TLS/SSL to prevent eavesdropping and man-in-the-middle attacks.

**Assessment Checklist:**
- [ ] Application Load Balancer uses HTTPS listeners
- [ ] CloudFront distributions use HTTPS (redirect HTTP to HTTPS)
- [ ] API Gateway uses TLS 1.2 or higher
- [ ] RDS connections use SSL/TLS
- [ ] ElastiCache connections use TLS (if supported)
- [ ] Application enforces HTTPS for all API endpoints
- [ ] TLS certificates managed via ACM (AWS Certificate Manager)
- [ ] TLS certificates auto-renewed

**MCP Validation:**
- Search: `aws_search_documentation("Application Load Balancer HTTPS listener")`
- Search: `aws_search_documentation("CloudFront HTTPS redirect")`
- Search: `aws_search_documentation("RDS SSL connection")`

**Risk Rating:**
- 🔴 RED: Application accepts HTTP traffic without HTTPS OR database connections not encrypted
- 🟡 YELLOW: HTTPS enabled but HTTP not redirected OR some services still using HTTP
- 🟢 GREEN: All traffic encrypted in transit with TLS 1.2+ and HTTP redirected to HTTPS

**Remediation Guidance:**
- Search: `aws_search_documentation("configure HTTPS ALB listener ACM")`
- Search: `aws_search_documentation("enforce SSL RDS connections")`
- Estimated Implementation Time: 1-2 hours

---

#### Control 2.5: VPC Endpoints for AWS Services

**Control Objective:** Use VPC endpoints to access AWS services privately without traversing the public internet, reducing attack surface and data transfer costs.

**Assessment Checklist:**
- [ ] VPC endpoint for S3 (Gateway endpoint)
- [ ] VPC endpoint for DynamoDB (Gateway endpoint)
- [ ] VPC endpoints for other AWS services used (Interface endpoints)
- [ ] VPC endpoint policies restrict access to specific resources
- [ ] Private subnets use VPC endpoints instead of NAT Gateway for AWS service access
- [ ] VPC endpoint DNS resolution enabled

**MCP Validation:**
- Search: `aws_search_documentation("VPC endpoints AWS services setup")`
- Search: `aws_search_documentation("VPC endpoint policies security")`
- Check: `aws_get_regional_availability(region="<your-region>", resource_type="api", filters=["EC2+CreateVpcEndpoint"])`

**Risk Rating:**
- 🔴 RED: No VPC endpoints configured, all AWS service traffic goes over internet
- 🟡 YELLOW: Some VPC endpoints configured but not for all services OR endpoint policies too permissive
- 🟢 GREEN: VPC endpoints configured for all AWS services with restrictive policies

**Remediation Guidance:**
- Search: `aws_search_documentation("create VPC endpoint S3 DynamoDB")`
- Estimated Implementation Time: 30-60 minutes

---

#### Control 2.6: Network Segmentation

**Control Objective:** Implement proper network segmentation using VPCs, subnets, and security groups to isolate workloads and limit blast radius.

**Assessment Checklist:**
- [ ] Resources deployed in VPC (not EC2-Classic)
- [ ] Public and private subnets properly separated
- [ ] Database tier in private subnets (no direct internet access)
- [ ] Application tier in private subnets behind load balancer
- [ ] Bastion hosts or Systems Manager Session Manager for administrative access (no direct SSH)
- [ ] Security groups configured for each tier (web, app, database)
- [ ] Network ACLs provide additional layer of defense
- [ ] VPC Flow Logs enabled for network traffic monitoring

**MCP Validation:**
- Search: `aws_search_documentation("VPC network segmentation best practices")`
- Search: `aws_search_documentation("VPC Flow Logs enable")`

**Risk Rating:**
- 🔴 RED: Resources in default VPC OR databases in public subnets OR no network segmentation
- 🟡 YELLOW: Custom VPC but weak segmentation OR VPC Flow Logs not enabled
- 🟢 GREEN: Proper network segmentation with defense in depth and flow logs enabled

**Remediation Guidance:**
- Search: `aws_search_documentation("migrate resources to custom VPC")`
- Search: `aws_search_documentation("create private subnet database tier")`
- Estimated Implementation Time: 4-8 hours (significant architectural change)

---

### Step 4: Monitoring & Compliance Controls Assessment

#### Control 3.1: Centralized Logging

**Control Objective:** Aggregate logs from all sources for security monitoring, troubleshooting, and compliance.

**Assessment Checklist:**
- [ ] Application logs sent to CloudWatch Logs
- [ ] VPC Flow Logs enabled and sent to CloudWatch Logs or S3
- [ ] Load balancer access logs enabled
- [ ] S3 access logs enabled for sensitive buckets
- [ ] Lambda function logs sent to CloudWatch Logs
- [ ] Log retention configured (minimum 90 days for security logs)
- [ ] Log aggregation to central S3 bucket for long-term storage
- [ ] Log analysis tools configured (CloudWatch Insights, Athena, or third-party SIEM)

**MCP Validation:**
- Search: `aws_search_documentation("centralized logging architecture AWS")`
- Search: `aws_search_documentation("CloudWatch Logs retention policy")`

**Risk Rating:**
- 🔴 RED: No centralized logging OR logs retained <30 days
- 🟡 YELLOW: Centralized logging exists but incomplete OR retention <90 days
- 🟢 GREEN: Comprehensive centralized logging with appropriate retention and analysis tools

**Remediation Guidance:**
- Search: `aws_search_documentation("enable VPC Flow Logs CloudWatch")`
- Search: `aws_search_documentation("ALB access logs S3")`
- Estimated Implementation Time: 2-3 hours

---

#### Control 3.2: Security Monitoring & Alerting

**Control Objective:** Implement real-time security monitoring and alerting for critical security events.

**Assessment Checklist:**
- [ ] CloudWatch alarms for root user login
- [ ] CloudWatch alarms for IAM policy changes
- [ ] CloudWatch alarms for security group changes
- [ ] CloudWatch alarms for CloudTrail configuration changes
- [ ] CloudWatch alarms for failed authentication attempts
- [ ] GuardDuty findings sent to SNS/email for high severity
- [ ] AWS Config rules for compliance monitoring
- [ ] Security Hub enabled for centralized security findings (optional but recommended)

**MCP Validation:**
- Search: `aws_search_documentation("CloudWatch alarms security events")`
- Search: `aws_search_documentation("AWS Config rules security compliance")`
- Check: `aws_get_regional_availability(region="<your-region>", resource_type="api", filters=["SecurityHub+EnableSecurityHub", "Config+PutConfigRule"])`

**Risk Rating:**
- 🔴 RED: No security monitoring or alerting configured
- 🟡 YELLOW: Some alarms configured but missing critical events OR alerts not actionable
- 🟢 GREEN: Comprehensive security monitoring with actionable alerts and response procedures

**Remediation Guidance:**
- Search: `aws_search_documentation("create CloudWatch alarm root user login")`
- Search: `aws_search_documentation("AWS Config rules security best practices")`
- Estimated Implementation Time: 2-4 hours

---

#### Control 3.3: Incident Response Preparation

**Control Objective:** Establish incident response procedures and tools to quickly detect, respond to, and recover from security incidents.

**Assessment Checklist:**
- [ ] Incident response plan documented
- [ ] Incident response team identified with contact information
- [ ] Incident response runbooks created for common scenarios
- [ ] AWS support plan appropriate for response time requirements (Business or Enterprise)
- [ ] Backup and recovery procedures tested
- [ ] Forensic investigation tools identified (EC2 snapshots, memory dumps, etc.)
- [ ] Communication plan for security incidents (internal and external)
- [ ] Incident response plan tested annually

**MCP Validation:**
- Search: `aws_search_documentation("AWS incident response best practices")`
- Search: `aws_search_documentation("security incident forensics AWS")`

**Risk Rating:**
- 🔴 RED: No incident response plan OR plan never tested
- 🟡 YELLOW: Incident response plan exists but incomplete OR not tested in >12 months
- 🟢 GREEN: Comprehensive incident response plan with regular testing and updates

**Remediation Guidance:**
- Search: `aws_search_documentation("incident response plan template")`
- Estimated Implementation Time: 4-8 hours (initial plan creation)

---

## Security Posture Scoring Methodology

### Overall Security Score Calculation

**Account-Level Controls (60% weight):**
- Control 1.1 (Root User Security): 15%
- Control 1.2 (IAM Password Policy): 5%
- Control 1.3 (CloudTrail Logging): 15%
- Control 1.4 (AWS Budget Alerts): 5%
- Control 1.5 (Amazon GuardDuty): 10%
- Control 1.6 (Trusted Advisor): 5%
- Control 1.7 (Account Contacts): 2.5%
- Control 1.8 (IAM User Management): 2.5%

**Workload-Level Controls (40% weight):**
- Control 2.1 (Secrets Management): 10%
- Control 2.2 (Resource Access): 7.5%
- Control 2.3 (Encryption at Rest): 10%
- Control 2.4 (Encryption in Transit): 5%
- Control 2.5 (VPC Endpoints): 2.5%
- Control 2.6 (Network Segmentation): 5%

**Monitoring & Compliance (Bonus Points):**
- Control 3.1 (Centralized Logging): +5%
- Control 3.2 (Security Monitoring): +5%
- Control 3.3 (Incident Response): +5%

**Scoring Scale:**
- 🔴 RED Control: 0 points
- 🟡 YELLOW Control: 50 points
- 🟢 GREEN Control: 100 points

**Overall Security Maturity Levels:**
- 0-40%: 🔴 **Critical Risk** - Not production-ready, immediate remediation required
- 41-60%: 🟡 **High Risk** - Significant gaps, production deployment not recommended
- 61-75%: 🟡 **Medium Risk** - Acceptable for early-stage startups, prioritize remediation
- 76-90%: 🟢 **Low Risk** - Production-ready with minor improvements needed
- 91-100%: 🟢 **Excellent** - Strong security posture, ready for enterprise customers

---

## Output Format

For each control category, provide:

### 1. Executive Summary Dashboard

Overall Security Score: [X]% - [Risk Level]

Account-Level Controls: [X]% ([Y] of [Z] controls GREEN) Workload-Level Controls: [X]% ([Y] of [Z] controls GREEN) Monitoring & Compliance: [X]% ([Y] of [Z] controls GREEN)

Production Readiness: [YES/NO] Investor Due Diligence Ready: [YES/NO] Compliance Framework Readiness: [SOC2: YES/NO, HIPAA: YES/NO, etc.]


### 2. Control-by-Control Assessment Report

For each control, provide:
- Control ID and Name
- Risk Rating (🔴/🟡/🟢)
- Current Status (Implemented / Partially Implemented / Not Implemented)
- Gap Analysis (what's missing)
- AWS Documentation References (URLs from MCP searches)
- Remediation Priority (P0 Critical /

¿Cómo se utiliza?

BETA
Required Access:
• AWS Knowledge MCP Server integration enabled in AI assistant
• AWS account access (read-only sufficient for assessment)
• Basic understanding of AWS services (EC2, RDS, S3, IAM)

Tools Required:
• AI assistant with MCP support
• Access to MCP tools: aws_search_documentation, aws_read_documentation, aws_get_regional_availability, aws_list_regions

SETUP INSTRUCTIONS
Step 1: Prepare Assessment Context
Gather the following information:
• Startup Profile: Company name, funding stage, industry, current scale
• AWS Environment: Services in use, account structure, team size
• Security Goals: Primary objectives, timeline, budget, compliance requirements
Step 2: Configure the Prompt
• Copy the complete prompt composition
• Replace all bracketed placeholders [LIKE_THIS] with your specific information
• Ensure all sections are completed, especially Current Environment Context
Step 3: Execute Assessment
• Submit configured prompt to AI assistant with MCP support
• Assistant will systematically assess each control using MCP tools
• Review output for completeness (Executive Summary, Control Reports, Remediation Roadmap)

Al usar estas petición, usted acepta el descargo de responsabilidad.