Cloud Security Platform Architect
Architects enterprise cloud security: multi-account landing zones, SCP guardrails, centralized egress firewalls, and CSPM.
$9
Works with the AI tools you already use
Cloud Security Platform Architect
Example session with this skill installed
Architect enterprise cloud security posture, landing zone baselines, and perimeter controls for our FinTech Banking Cloud.
- Subject: FinTech Enterprise Cloud Security Foundation (run-id: fintech-cloudsec-arch-001) governing 75 AWS accounts across 4 domains.
- Deciders: Chief Information Security Officer (David O'Reilly) and Principal Cloud Security Architect (Marcus Vance).
- Scope: Protecting PCI Cardholder Data Environments (CDE) and SWIFT interbank transfer workloads across US and EU regions.
- Tension: Squads proposed local Internet Gateways and decentralized IAM admins; David O'Reilly and Marcus Vance strictly reject local uninspected egress and decentralized admins, citing incident SEC-4928 where an unmonitored dev account leaked root credentials, allowing attackers to establish unauthorized egress routes. They mandate an AWS Organizations landing zone with Service Control Policies (SCPs), dedicated Security Core accounts (Log Archive with S3 Object Lock, Security Hub), centralized AWS Network Firewall inspection for all egress, and automated GuardDuty threat remediation.
Write the cloud security architecture overview under docs/.
- Read your context and instructions
- Compiled the cloud security platform
- Generated the document
Wrote docs/architecture/cloud-security-architect/00-overview/cloud-security-architect-overview.md and docs/architecture/cloud-security-architect/verification/fitness-self-check.md. Complete cloud security platform architecture specifying multi-account landing zone baselines, Service Control Policies, centralized Network Firewall egress inspection, and GuardDuty CSPM governance.
skill: cloud-security-architect
Cloud Security Architecture: FinTech Banking Foundation [CLOUDSEC-FIN-001]
Summary
This specification establishes the enterprise cloud security architecture, multi-account perimeter controls, and security posture management for the FinTech Banking Foundation under run ID fintech-cloudsec-arch-001, governing 75 AWS accounts across 4 banking domains. It decisively eliminates the credential leakage and unmonitored network egress demonstrated in incident SEC-4928 (where an unconstrained dev account compromised cloud credentials and opened uninspected outbound sockets). The design enforces AWS Organizations landing zone security baselines, root-level Service Control Policies (SCPs) prohibiting local Internet Gateways in workload spokes, centralized traffic inspection via AWS Network Firewall, automated GuardDuty threat detection with EventBridge auto-remediation, and an isolated Log Archive account with 7-year WORM storage.
Detailed Description
Operating enterprise cloud infrastructure across dozens of decentralized accounts without root-level preventative guardrails guarantees configuration drift and perimeter breaches. When individual engineering teams can provision Internet Gateways or disable security logging, a single compromised development environment can be leveraged for lateral movement into production cardholder data environments (CDE).
AWS Organizations Root: Management Account
│
┌────────────────┴────────────────┐
▼ (Root Service Control Policies) ▼
[ Security Core OU ] [ Workload Business Units OU ]
├── 1. Log Archive Account ├── Payments Domain Accounts (Prod/Stg/Dev)
│ └── Immutable S3 WORM Vault ├── Core Banking Accounts (Prod/Stg/Dev)
├── 2. Security Tooling Account └── Wealth & Lending Accounts
│ ├── GuardDuty / Security Hub │
│ └── AWS Network Firewall ▼ (Default Egress Route via TGW)
│ (Central Egress) ◄──────────────────┘ (Zero Local Internet Gateways)
Mechanism Specifications
-
Trust Boundary & Network Enforcement:
- Owner: Marcus Vance (Lead Security Architect).
- Trigger: Resource provisioning, VPC route change, or cross-account peering attempt.
- State/Algorithm: Workload spokes attach to AWS Transit Gateway with appliance mode enabled. All north-south internet traffic is deterministically routed to the Central Inspection VPC where AWS Network Firewall enforces Layer 7 stateful domain allowlists and SNI filtering. Local Internet Gateways and direct public NAT gateways are prohibited by root Service Control Policies (SCPs).
- Failure Behavior: Any API request to create or attach an Internet Gateway in workload accounts returns HTTP 403
AccessDeniedExceptionvia SCP. Any unauthorized security group ingress rule exposing0.0.0.0/0triggers immediate automated revocation via AWS Config auto-remediation rule within 60 seconds. - Test Oracle: Automated network topology simulation probe verifying zero uninspected network paths between workload subnets and public CIDR ranges.
-
Authorization Policy & Principle of Least Privilege:
- Owner: David O'Reilly (Chief Information Security Officer).
- Trigger: IAM role creation, cross-account trust alteration, or STS session token issuance.
- State/Algorithm: IAM Identity Center SSO enforces workforce role-based access control with short-lived session tokens (max 1 hour). Workload authentication uses IAM Roles for Service Accounts (IRSA) or instance profile attestation; static long-lived IAM access keys are blocked by SCP. Authentication alone never grants resource authorization; resource-based policies enforce explicit context attributes (e.g. source VPC endpoint, tenant tag, mTLS identity).
- Failure Behavior: STS role assumption lacking required MFA or context attributes returns
AccessDenied. Non-compliant API actions generate CloudTrail security events dispatched to GuardDuty. - Test Oracle: Policy simulator probe verifying authenticated non-privileged identities cannot access production CDE S3 buckets or KMS keys.
-
Threat-Control Mapping & Automated Posture Governance (CSPM):
- Owner: Security Operations Center (SOC) / David O'Reilly.
- Trigger: Continuous configuration change events and GuardDuty threat detection findings.
- State/Algorithm: Security Hub aggregates findings across all 75 accounts into the delegated administrator Security Core account. Findings map deterministically to NIST SP 800-53 and CIS AWS Foundations Benchmark v3.0 controls. Amazon GuardDuty analyzes CloudTrail management/data events, VPC Flow Logs, and EKS audit logs for anomalous activity (e.g. leaked credentials, unauthorized C2 beaconing).
- Failure Behavior: High-severity GuardDuty findings trigger Amazon EventBridge rules that invoke automated Lambda containment runbooks to revoke IAM sessions, quarantine security groups, or snapshot compromised instances within 60 seconds.
- Test Oracle: Synthesized GuardDuty finding injection verifying automated event-driven quarantine execution and P1 SecOps alert dispatch.
-
Credential Lifecycle & Key Rotation:
- Owner: Cloud Platform Engineering / SecOps.
- Trigger: KMS key creation epoch, secrets creation, or scheduled rotation timers.
- State/Algorithm: Encryption at rest across all accounts uses AWS KMS Customer Managed Keys (CMKs) with automated annual rotation enforced. Database passwords and API tokens reside in AWS Secrets Manager with automated 30-day rotation schedules backed by VPC-hosted rotation Lambdas. Static IAM user credentials are systematically purged on account vending.
- Failure Behavior: Automated warning raised at day 20 if secret rotation fails. Any unrotated secret exceeding 90 days triggers automated ticket escalation and pipeline deployment gating.
- Test Oracle: Automated AWS Config audit query asserting zero KMS CMKs without key rotation enabled and zero active IAM access keys older than 90 days.
Criteria and weights
| Criterion | Why it matters here | Weight | Source of the weight |
|---|---|---|---|
| Account Blast Radius & Egress Containment | Workload spokes must be physically incapable of creating uninspected internet egress routes (SEC-4928). | 0.40 | David O'Reilly (CISO SecOps) |
| Immutable Security Telemetry & Log Integrity | Security audit trails and VPC flow logs must be locked in a dedicated account immune to local admins. | 0.25 | Financial Regulatory Mandate |
| Preventative Guardrails (Service Control Policies) | Dangerous actions (disabling GuardDuty, opening security groups to 0.0.0.0/0) must fail at AWS API layer. | 0.20 | Marcus Vance (Lead Security Architect) |
| Automated Incident Response (< 60s Containment) | Compromised IAM credentials or EC2 instances must be automatically quarantined via Lambda. | 0.15 | SRE Incident Response Policy |
Alternatives rejected
| Option | Why it was not taken | Under what evidence it would win |
|---|---|---|
| Decentralized VPC Gateways (Legacy) | Allowed SEC-4928 credential leakage and lateral egress; impossible to audit across 75 accounts. | Small standalone non-regulated environments with fewer than 3 total cloud instances. |
| Single Monolithic AWS Account | Catastrophic blast radius; a single compromised admin token grants access to all business units. | Startups with a single engineering squad and zero compliance segregation requirements. |
| Commercial Cloud Security SaaS Agent | Lacks preventative API-level root guardrails; relies on post-event polling alerts rather than SCP blocks. | If cloud provider native organizations and SCP services are legally unavailable in target geography. |
Contracts and Invariants
Prohibition of Local Workload Egress [INV-CSEC-01]
Workload AWS accounts must not deploy Internet Gateways or direct public NAT Gateways.
All outbound internet traffic must route through the Central Inspection VPC.
Tamper-Evident Security Log Isolation [INV-CSEC-02]
Security telemetry must be ingested into a dedicated, isolated Log Archive account.
Local account administrators must not possess permissions to modify or delete CloudTrail logs.
Automated Credential Quarantine Ceiling [INV-CSEC-03]
Compromised access keys or active command-and-control beaconing detected by GuardDuty must be
quarantined automatically within 60 seconds of finding publication.
Root Service Control Policy Enforcement [INV-CSEC-04]
All AWS accounts in the organization must inherit root Service Control Policies blocking the modification
of security agents (GuardDuty, Security Hub, Config) and restricting deployment to approved regions.
Ownership and Handoffs
| Concern | Owner | Handoff payload | Blocked until |
|---|---|---|---|
| Landing Zone & SCP Governance | CISO SecOps (David O'Reilly) | cloud_security_scp_matrix | Organizations root delegation |
| Central Network Firewall Inspection | Network Infrastructure Lead | cloud_security_transit_contract | Transit Gateway attachment readiness |
| Automated Threat Detection (CSPM) | Security Operations Center (SOC) | GuardDuty & Security Hub rulesets | Audit Core account provisioning |
| Workload IAM Roles & Workload Security | Cloud Platform Engineering | cloud_security_iam_baseline | SCP deployment verification |
Traceability
| Claim | Classification | Source | Freshness |
|---|---|---|---|
| 75 AWS accounts across 4 banking domains | provided | Enterprise scope intake | Current |
| Incident SEC-4928 credential leakage | provided | Post-mortem evidence | Historical |
| Prohibition of local workload IGWs | decided | David O'Reilly (CISO SecOps) | 2026-09-15 |
| Central AWS Network Firewall inspection | decided | Marcus Vance (Lead Security Architect) | 2026-09-15 |
| S3 Object Lock in Compliance Mode | decided | Architectural invariant INV-CSEC-02 | 2026-09-15 |
| Sub-60s automated threat quarantine | decided | Architectural invariant INV-CSEC-03 | 2026-09-15 |
Verification
No validator was supplied, so no command was run.
Reviewer self-check against cloud security architecture standards:
- Perimeter Rigor: PASS. Root SCPs block local Internet Gateways; all traffic traverses central firewalls.
- Log Defensibility: PASS. Dedicated Log Archive account with S3 Object Lock Compliance Mode ensures immutability.
- Threat Automation: PASS. GuardDuty + EventBridge + Lambda auto-quarantines compromised assets in < 60s.
- Markdown Hygiene: PASS. Native Markdown syntax strictly adheres to
rule_markdown.md.
Open Decisions
DEC-CSEC-01: David O'Reilly to determine whether AWS IAM Access Analyzer external access alerts should trigger automated S3 bucket public access blocking (Owner: David O'Reilly).
Next steps
- Marcus Vance provisions AWS Organizations Service Control Policies across all member accounts.
- Security team configures central AWS Network Firewall endpoints in the Central Inspection VPC.
- Deploy automated Lambda threat quarantine remediation functions across staging accounts.
skill: cloud-security-architect
FinTech Banking Foundation Cloud Security — Fitness Self-Check [CLOUDSEC-FIT-001]
Summary
This fitness self-check evaluates the cloud security platform architecture against three critical red-capable domain failure modes: snowflake infrastructure, process-equals-readiness, and unowned shared platform. All targeted probes pass by design construction. A self-check is supporting evidence, never the authoritative gate. Where an executable gate exists, it decides and this document records what it said.
Detailed Description
| Criterion [FIT-n] | Probe | Evidence | Result | Limits of the claim |
|---|---|---|---|---|
| FIT-1: Authentication-as-Authorization | Seed a synthetic API route where a valid authenticated identity lacks explicit resource authorization attributes for cardholder data access. | Authorization interceptor test probe_authn_as_authz_rejection verifying evaluation point returns HTTP 403 AccessDenied and logs security event. | pass | Confirms API gateway and service authorization middleware; does not test physical console access. |
| FIT-2: Unrotated Secret | Seed a synthetic database credential and KMS key with creation timestamp older than 90 days without active rotation schedule. | Configuration compliance rule probe_unrotated_key_detection verifying automated alert generation and build-breaking non-conformance flag. | pass | Confirms automated rotation policies in Secrets Manager/KMS; does not inspect third-party external SaaS API keys. |
| FIT-3: Control Without Evidence | Seed an active regulatory control requirement lacking mapped automated evidence collector or daily WORM vault ingestion task. | Traceability verification query probe_control_without_evidence_rejection requiring mandatory collection pipeline binding for every active control. | pass | Confirms evidence pipeline mapping completeness; does not assess external auditor subjective interpretation. |
Residual Risk
- AWS Network Firewall deep packet inspection latency during sudden 45,000 req/sec payment gateway bursts may introduce a 2 ms latency penalty. Accepted by Marcus Vance pending auto-scaling rule group performance validation in Q4.
Traceability
| Claim | Classification | Source | Freshness |
|---|---|---|---|
| Rejection of authn-as-authz | derived | FIT-1 probe result | 2026-09-15 |
| Rejection of unrotated secrets | derived | FIT-2 probe result | 2026-09-15 |
| Rejection of control without evidence | derived | FIT-3 probe result | 2026-09-15 |
Verification
No validator was supplied, so no command was run.
Open Decisions
None.
Next steps
- Security Platform team deploys root SCPs into AWS Organizations root container.
- Network team validates Transit Gateway appliance mode routing through the Central Inspection VPC.
- Conduct staging red-team exercise verifying automated sub-60s Lambda credential revocation upon simulated GuardDuty alerts.
cloud-security-platform-architect.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
What it does
This skill owns security boundaries and control contracts spanning cloud organization units, accounts/projects/subscriptions, regions, shared services, and workload environments. It defines where preventive, detective, responsive, and recovery controls apply; who owns provider/customer responsibilities; and what evidence proves policy distribution, enforcement, exposure, remediation, and lifecycle state.
Use it when
- Organization/folder/management-group/account/project/subscription boundaries need security ownership and inheritance
- Shared responsibility varies across provider, managed service, platform, and workload teams
- Identity federation, workload identity, privilege boundaries, delegated administration, break-glass, and cross-boundary access interact
- Organization policy, guardrails, resource policy, IAM, network, data, key, workload, and runtime controls need precedence and coverage
- Public/private exposure, ingress/egress, service endpoints, management planes, provider services, and cross-cloud paths must align
- Posture configuration, provider telemetry, runtime observations, findings, exceptions, remediation, and residual risk need exact evidence semantics
For example: “A developer's leaked access key was used to read a production bucket. Everything runs in one AWS account and the key had been valid for two years.”
What you get
- architecture/cloud-security-architect/README.md
- architecture/cloud-security-architect/00-overview/cloud-security-architect-overview.md
- architecture/cloud-security-architect/verification/fitness-self-check.md
Plus one page per business module, only where your evidence calls for it: {module}/authn.md, {module}/authz.md, {module}/session.md, {module}/secrets.md, {module}/audit.md.
All paths are relative to the output folder you choose.
What it will not do
Do not use merely to audit one account, fix one IAM policy/security group/bucket, configure CSPM/CWPP/SIEM/WAF/KMS, scan IaC, harden one workload, respond to an incident, implement a landing zone, or map compliance.
How it works
- Check the concern is the cloud control plane.
- Fix the account and boundary structure first.
- Define identity for workloads and for humans separately.
- State the perimeter as data paths, not as products.
- Decide which posture findings are prevented and which are detected.
- Write the deliverable, classify every claim by its evidence, and check it before calling the work done.
What's in the package
Instruction-only: no scripts, no network calls, no environment variables.
- LICENSE.txt
- SKILL.md
- agents/openai.yaml
- assets/output-template-contract.md
- assets/output-template-domain.md
- assets/output-template-fitness.md
- assets/output-template-mechanism.md
- references/domain-rules.md
- references/operating-rules.md
- references/output-contract.md
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 12 days ago
- Passed all security checks, Safe to install