azure-security-analyzer

name	azure-security-analyzer
description	Analyze Azure resource configurations against security best practices using Azure MCP bestpractices service. Produces per-resource security assessment with severity ratings and recommendations. Use during template generation before deployment confirmation.
metadata	{"argument-hint":"Resource types and their configurations from the ARM template","user-invocable":true}

Azure Security Analyzer

Analyze Azure resource configurations against Microsoft security best practices and produce a per-resource security assessment report.

When to Use

• During template generation (invoked by the template generator before deployment confirmation)
• To audit an existing ARM template for security gaps
• When user asks "is this secure?" or "check security" for a deployment
• Post-deployment security review

Verification Integrity Rules (CRITICAL)

Every claim in the security report MUST be verifiable against the ARM template. Never fabricate, assume, or misrepresent security status.

Rule 1: Cite Exact Evidence

Every "✅ Applied" status MUST cite the exact ARM template property path and its value that proves the control is in place. If you cannot point to a specific property in the template JSON, you cannot mark it as applied.

# ✅ CORRECT — cites exact property
| HTTPS-only | 🔴 Critical | ✅ Applied | `properties.httpsOnly: true` | Explicitly set in template |

# ❌ WRONG — no evidence from template
| Disk encryption | 🔴 Critical | ✅ Applied | `managedDisk.storageAccountType` | This property is the performance tier, NOT encryption |

Rule 2: Distinguish Explicit Config vs Platform Defaults

Azure provides some security controls by default (e.g., SSE at rest on managed disks). These are NOT the same as explicitly configured controls.

Status	Icon	Meaning	When to Use
✅ Applied	✅	Explicitly configured in the ARM template	Property exists in template JSON with secure value
🔄 Platform Default	🔄	Azure provides this automatically, NOT in template	Control exists due to Azure platform behavior, not template config
⚠️ Not applied	⚠️	Control is missing and should be considered	Property absent from template, no platform default covers it
❌ Misconfigured	❌	Property exists but set to an insecure value	Property in template with wrong/insecure value

# ✅ CORRECT — distinguishes platform default from explicit config
| SSE at rest (managed disks) | 🟡 Medium | 🔄 Platform Default | Not in template | Azure encrypts all managed disks at rest with platform-managed keys automatically |
| Encryption at host (ADE) | 🟡 Medium | ⚠️ Not applied | `securityProfile.encryptionAtHost` absent | Not enabled — would encrypt temp disks and caches too |

# ❌ WRONG — claims explicit config for a platform default
| Managed disk encryption | 🔴 Critical | ✅ Applied | `managedDisk.storageAccountType` | WRONG: storageAccountType is performance tier, not encryption |

Rule 3: Never Use Misleading Framing

Describe security status accurately and literally. Do not soften or reframe risks.

# ❌ WRONG — misleading framing
| No open SSH to internet | 🔴 Critical | ✅ Applied |
# This is misleading: the VM HAS a public IP with port 22 open.
# IP-restriction reduces risk but port 22 IS internet-reachable.

# ✅ CORRECT — accurate framing
| SSH not open to 0.0.0.0/0 | 🔴 Critical | ✅ Applied | `sourceAddressPrefix: 175.x.x.x/32` | Restricted to single IP |
| VM is internet-facing (public IP + port 22) | 🟠 High | ⚠️ Risk accepted | Public IP attached to NIC | Port 22 reachable from internet (IP-restricted). Safer: Azure Bastion |

Specific rules:

• If a VM has a public IP → it IS internet-facing. Always flag this.
• If any port is open, even IP-restricted → state the port IS open and note the restriction as mitigation, not as "closed."
• If encryption is platform-default → say "platform default", not "applied."
• If a property is absent from the template → say "not configured", not "applied" based on assumptions.

Rule 4: Verify Before Reporting

Before generating the security report, perform this verification checklist:

1. For each "✅ Applied" entry: Search the ARM template JSON for the cited property. If not found → change status.
2. For each security claim: Confirm the ARM property cited actually controls what you claim. (e.g., storageAccountType is NOT encryption)
3. For network exposure: Check if any publicIPAddress resource is attached to a NIC. If yes → VM is internet-facing, period.
4. For encryption claims: Distinguish between SSE (automatic), ADE (explicit extension), and encryption at host (explicit VM property).
5. Cross-check property paths: Use the correct ARM template property paths, not invented or approximate ones.

Rule 5: When Uncertain, Mark as Unknown

If you cannot determine a security status with certainty:

| {check} | {severity} | ❓ Unknown | {property} | Unable to verify — property path unclear or resource type not in checklist |

Never guess. Never fabricate. When in doubt, flag it.

Procedure

1. Extract Resources from Template

Parse the ARM template or requirements to identify all resources and their configurations:

Input: ARM template JSON or resource configuration list

Extract for each resource:
- Resource type (e.g., Microsoft.Storage/storageAccounts)
- Resource name
- All security-relevant properties
- Current configuration values

2. Query Azure MCP Best Practices (Per Resource)

For EACH resource type, query the Azure MCP bestpractices service to get security recommendations:

Tool: mcp_azure_mcp_search
Intent: "bestpractices {resource-type}"

Examples:
- "bestpractices Microsoft.Storage/storageAccounts"
- "bestpractices Microsoft.Web/sites"
- "bestpractices Microsoft.Sql/servers"
- "bestpractices Microsoft.DocumentDB/databaseAccounts"
- "bestpractices Microsoft.KeyVault/vaults"
- "bestpractices Microsoft.ContainerApp/containerApps"

Parse the MCP response for:

• Security recommendations with severity levels
• Configuration best practices
• Compliance alignment notes (CIS, SOC2, PCI-DSS)
• Microsoft Defender for Cloud recommendations

3. Check Resource-Specific Security Properties

Cross-reference MCP recommendations against the template configuration.

Storage Accounts (Microsoft.Storage/storageAccounts):

#	Check	Property	Severity	Secure Value
1	HTTPS-only transfer	supportsHttpsTrafficOnly	🔴 Critical	true
2	TLS 1.2 minimum	minimumTlsVersion	🔴 Critical	TLS1_2
3	Disable public blob access	allowBlobPublicAccess	🟠 High	false
4	Blob soft delete	deleteRetentionPolicy.enabled	🟠 High	true
5	Container soft delete	containerDeleteRetentionPolicy.enabled	🟡 Medium	true
6	Disable shared key access	allowSharedKeyAccess	� High	false
7	Network rules / firewall	networkAcls.defaultAction	🟡 Medium	Deny (prod)
8	Private endpoint	Private endpoint resource	🟡 Medium	Configured (prod)
9	Infrastructure encryption	encryption.requireInfrastructureEncryption	🔵 Low	true
10	Immutability policy	immutableStorageWithVersioning	🔵 Low	Enabled (compliance)

Function Apps / App Services (Microsoft.Web/sites):

#	Check	Property	Severity	Secure Value
1	HTTPS-only	httpsOnly	🔴 Critical	true
2	TLS 1.2 minimum	siteConfig.minTlsVersion	🔴 Critical	1.2
3	Managed identity	identity.type	🔴 Critical	SystemAssigned
4	Identity-based storage access	AzureWebJobsStorage__accountName (not AzureWebJobsStorage)	🔴 Critical	Identity-based
5	RBAC for storage	Role assignments for MI → Storage	🔴 Critical	Storage Blob Data Owner + Storage Account Contributor
6	FTP disabled	siteConfig.ftpsState	🟠 High	Disabled
7	Remote debugging off	siteConfig.remoteDebuggingEnabled	🟠 High	false
8	Latest runtime version	siteConfig.linuxFxVersion or netFrameworkVersion	🟠 High	Latest stable
9	App Insights connected	APPINSIGHTS_INSTRUMENTATIONKEY in appSettings	🟡 Medium	Set
8	Health check enabled	siteConfig.healthCheckPath	🟡 Medium	Set
9	CORS not wildcard	siteConfig.cors.allowedOrigins	🟡 Medium	Not *
10	VNet integration	virtualNetworkSubnetId	🟡 Medium	Configured (prod)
11	IP restrictions	siteConfig.ipSecurityRestrictions	🔵 Low	Configured
12	HTTP/2 enabled	siteConfig.http20Enabled	🔵 Low	true

SQL Servers (Microsoft.Sql/servers):

#	Check	Property	Severity	Secure Value
1	TDE enabled	transparentDataEncryption.status	🔴 Critical	Enabled
2	AAD-only auth	administrators.azureADOnlyAuthentication	🔴 Critical	true
3	Minimal TLS 1.2	minimalTlsVersion	🔴 Critical	1.2
4	Auditing enabled	auditingSettings.state	🟠 High	Enabled
5	Advanced threat protection	securityAlertPolicies.state	🟠 High	Enabled
6	Firewall rules restrictive	firewallRules	🟠 High	No 0.0.0.0/0 (prod)
7	Vulnerability assessment	vulnerabilityAssessments	🟡 Medium	Enabled
8	Private endpoint	Private endpoint resource	🟡 Medium	Configured (prod)
9	Long-term backup retention	backupLongTermRetentionPolicies	🟡 Medium	Configured
10	Connection policy	connectionPolicies.connectionType	🔵 Low	Redirect

Cosmos DB (Microsoft.DocumentDB/databaseAccounts):

#	Check	Property	Severity	Secure Value
1	Firewall configured	ipRules or virtualNetworkRules	🔴 Critical	Not empty
2	Disable key-based metadata writes	disableKeyBasedMetadataWriteAccess	🟠 High	true
3	RBAC-based access	disableLocalAuth	🟠 High	true (prefer RBAC)
4	Continuous backup	backupPolicy.type	🟡 Medium	Continuous
5	Customer-managed keys	keyVaultKeyUri	🟡 Medium	Set (prod)
6	Diagnostic logging	Diagnostic setting resource	🟡 Medium	Enabled
7	Private endpoint	Private endpoint resource	🟡 Medium	Configured (prod)

Key Vault (Microsoft.KeyVault/vaults):

#	Check	Property	Severity	Secure Value
1	Soft delete enabled	enableSoftDelete	🔴 Critical	true (now default)
2	Purge protection	enablePurgeProtection	🔴 Critical	true
3	RBAC authorization	enableRbacAuthorization	🟠 High	true
4	Network rules	networkAcls.defaultAction	🟡 Medium	Deny (prod)
5	Private endpoint	Private endpoint resource	🟡 Medium	Configured (prod)
6	Diagnostic logging	Diagnostic setting resource	🟡 Medium	Enabled

Container Apps (Microsoft.App/containerApps):

#	Check	Property	Severity	Secure Value
1	Ingress HTTPS only	ingress.transport	🔴 Critical	http with allowInsecure: false
2	Managed identity	identity.type	🟠 High	SystemAssigned
3	Min replicas > 0 (prod)	scale.minReplicas	🟡 Medium	> 0 for prod
4	VNet integration	Container Apps Environment	🟡 Medium	Configured
5	Secret management	secrets with Key Vault refs	🟡 Medium	Key Vault references

Virtual Machines (Microsoft.Compute/virtualMachines):

#	Check	Property	Severity	Secure Value
1	Password authentication disabled	osProfile.linuxConfiguration.disablePasswordAuthentication	🔴 Critical	true
2	SSH key configured	osProfile.linuxConfiguration.ssh.publicKeys	🔴 Critical	At least one key
3	Internet exposure (public IP)	Check if NIC references publicIPAddress	🟠 High	No public IP (use Bastion). If public IP present → flag as internet-facing
4	NSG SSH not open to 0.0.0.0/0	NSG securityRules[].sourceAddressPrefix for port 22	🔴 Critical	Not * or Internet — must be specific IP/CIDR
5	Encryption at host	securityProfile.encryptionAtHost	🟡 Medium	true (encrypts temp disks + caches)
6	Azure Disk Encryption (ADE)	ADE VM extension resource	🟡 Medium	Extension present
7	SSE at rest (managed disks)	Platform default — NOT a template property	🔵 Info	🔄 Platform Default — Azure encrypts all managed disks automatically. Do NOT mark as "✅ Applied"
8	Boot diagnostics	diagnosticsProfile.bootDiagnostics.enabled	🟡 Medium	true
9	Managed identity	identity.type	🟡 Medium	SystemAssigned (if VM accesses Azure resources)
10	Automatic OS updates	osProfile.linuxConfiguration.patchSettings.patchMode	🔵 Low	AutomaticByPlatform
11	Trusted launch	securityProfile.securityType	🟡 Medium	TrustedLaunch with vTPM and secure boot

⚠️ VM Internet Exposure Rule: If ANY Microsoft.Network/publicIPAddresses resource is attached to the VM's NIC, the VM IS internet-facing. Always flag this explicitly, even if NSG rules restrict source IPs. An IP-restricted port is still internet-reachable — the restriction is a mitigation, not a closure.

Network Security Groups (Microsoft.Network/networkSecurityGroups):

#	Check	Property	Severity	Secure Value
1	No rules with source * or Internet on management ports (22, 3389)	securityRules[].sourceAddressPrefix	🔴 Critical	Specific IP/CIDR only
2	Explicit deny-all inbound rule	Custom deny rule at high priority	🟠 High	Present
3	No overly permissive rules (* destination port)	securityRules[].destinationPortRange	🟠 High	Specific ports only
4	NSG flow logs	Separate flow log resource	🟡 Medium	Enabled

3.5 Verify All Findings (MANDATORY)

Before classifying posture, run the verification checklist from the Verification Integrity Rules section:

1. Re-read the ARM template JSON.
2. For every "✅ Applied" entry in your draft report, search for the exact property path and confirm the value.
3. For every network-related check, confirm whether a publicIPAddress resource exists and which ports are open.
4. For every encryption check, confirm whether the control is explicit (in template) or platform-default (Azure automatic).
5. Remove or downgrade any entry that fails verification.
6. Ensure no check uses misleading framing (see Rule 3).

If you find an error during verification: Fix it immediately. Do not present unverified findings to the user.

4. Classify Overall Security Posture & Gate Status

Based on the per-resource analysis, calculate an overall score AND the Security Gate status:

Security Gate (BLOCKING):
- ALL 🔴 Critical checks MUST pass → if any fail, gate = 🔴 BLOCKED
- ALL 🟠 High checks MUST pass → if any fail, gate = 🔴 BLOCKED  
- If all Critical + High pass → gate = 🟢 PASSED
- 🟡 Medium and 🔵 Low do NOT block deployment

"Pass" means status is ✅ Applied or 🔄 Platform Default.
"Fail" means status is ⚠️ Not applied, ❌ Misconfigured, or ❓ Unknown.

Overall Posture (informational — does NOT control gating):
- **EXCELLENT** — All critical+high passed, > 75% medium passed
- **GOOD** — All critical+high passed, > 50% medium passed
- **ACCEPTABLE** — All critical passed, some high missing
- **NEEDS ATTENTION** — Some high-severity checks failed
- **NEEDS IMMEDIATE ATTENTION** — Critical checks failed (DO NOT deploy without fixing)

The Security Gate is the authoritative deployment-blocking mechanism. The posture label is informational only. A posture of "ACCEPTABLE" still results in 🔴 BLOCKED if any High checks fail.

5. Generate Security Report

Produce a structured security assessment:

# Security Best Practices Analysis

**Deployment ID:** {deployment-id}
**Analyzed:** {timestamp}
**Source:** Azure MCP bestpractices service
**Resources Analyzed:** {count}

## Overall Security Posture: {EXCELLENT|GOOD|ACCEPTABLE|NEEDS ATTENTION}

**Summary:**
- 🔴 Critical: {passed}/{total} passed
- 🟠 High: {passed}/{total} passed
- 🟡 Medium: {passed}/{total} passed
- 🔵 Low: {passed}/{total} passed

## Per-Resource Assessment

### {Resource Type}: {Resource Name}

| # | Recommendation | Severity | Status | Property | Evidence | Notes |
|---|---------------|----------|--------|----------|----------|-------|
| 1 | {check name} | {severity} | ✅ Applied | {exact property path} | {actual value from template} | {note} |
| 2 | {check name} | {severity} | 🔄 Platform Default | {property} | Not in template | {Azure provides this automatically} |
| 3 | {check name} | {severity} | ⚠️ Not applied | {property} | Absent | {recommendation} |

**Score: {applied}/{total}** ({percentage}%)

{Repeat for each resource}

## Recommendations

### Must Fix (before deployment)
{List any failed critical checks — these should block deployment}

### Should Fix (strongly recommended)
{List any failed high checks — warn user but don't block}

### Consider (best practice)
{List failed medium/low checks — informational}

## Environment-Specific Notes

**For production deployments:**
- Private endpoints should be configured for all data resources
- Network rules should default to Deny
- VNet integration recommended for compute resources
- Consider customer-managed keys for encryption

**For dev/staging:**
- Network rules can use Allow for easier development
- Private endpoints are optional
- Shared key access acceptable for development

6. Return Results with Gate Status

Return the security report to the calling agent (template generator) with two key outputs:

1. Security Report — Full per-resource assessment (saved to security-analysis.md)
2. Security Gate Status — 🟢 PASSED or 🔴 BLOCKED with list of blocking findings

The gate status MUST be prominently displayed at the top of the report:

## Security Gate: 🟢 PASSED
All Critical and High security checks pass. Deployment may proceed.

or:

## Security Gate: 🔴 BLOCKED
Deployment cannot proceed. The following checks must be resolved:

| # | Check | Severity | Resource | Status | Required Fix |
|---|-------|----------|----------|--------|--------------|
| 1 | {check} | 🔴 Critical | {resource} | ⚠️ Not applied | {fix} |

Also save the report to .azure/deployments/$DEPLOYMENT_ID/security-analysis.md.

Save the gate result to .azure/deployments/$DEPLOYMENT_ID/security-gate.json:

{
  "gate": "PASSED",
  "blockingFindings": [],
  "criticalTotal": 4,
  "criticalPassed": 4,
  "highTotal": 6,
  "highPassed": 6,
  "timestamp": "2026-02-19T10:00:00Z"
}

Severity Classification

Severity	Icon	Meaning	Action
Critical	🔴	Security vulnerability if missing	Must fix before deployment
High	🟠	Strongly recommended by Microsoft	Should fix, warn user
Medium	🟡	Best practice, may have trade-offs	Recommend, user decides
Low	🔵	Nice to have, optional	Informational only

Environment Sensitivity

Some checks are environment-dependent:

Check	Dev	Staging	Prod
Private endpoints	🔵 Optional	🟡 Recommended	🟠 Required
Network firewall (Deny)	🔵 Optional	🟡 Recommended	🟠 Required
VNet integration	🔵 Optional	🟡 Recommended	🟠 Required
Customer-managed keys	🔵 Optional	🔵 Optional	🟡 Recommended
IP restrictions	🔵 Optional	🟡 Recommended	🟠 Required
Shared key access	🟡 Acceptable	🟡 Restricted	🔴 Disabled

When analyzing, adjust severity ratings based on the target environment from the requirements.

Usage

Invoked by template generator:

/azure-security-analyzer

Input: ARM template resources and target environment
Output: Per-resource security assessment with severity ratings

Manual invocation:

User: /azure-security-analyzer

Agent: Provide the ARM template or resource configurations to analyze.

User: [pastes template or points to deployment]

Agent: Analyzing 3 resources against Azure best practices...

[Produces security report]

Post-deployment audit:

User: /azure-security-analyzer --deployment-id deploy-20260218-193500

Agent: Loading template from .azure/deployments/deploy-20260218-193500/template.json...

[Analyzes and produces report]

Error Handling

If MCP bestpractices service is unavailable:

1. Fall back to the built-in security checklists defined in this skill
2. Note in the report: "Analysis based on built-in checklists (MCP bestpractices unavailable)"
3. Still produce a complete report — the checklists above are comprehensive

If a resource type is not in the checklists:

1. Query MCP bestpractices for the specific resource type
2. If no results, note: "No security checklist available for {resource-type}"
3. Apply generic checks: tags, diagnostic logging, encryption