Security audit

azure-cloud-architect-free

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a read-only Azure inventory helper, but it asks for broad local command authority and gives inconsistent guidance around Azure login, subscription context, and security-scope boundaries.

Review this skill before installing if you use Azure accounts with production access. Only run it with a least-privilege Azure identity, manually approve any az login or az account set action, and treat the security/compliance claims as unreliable unless the publisher tightens the scope and removes the unnecessary write authority.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The skill repeatedly markets itself as read-only, but it instructs users to perform interactive authentication and session-establishing actions such as `az login --use-device-code`. Even though login is not a destructive cloud-resource change, it is still a state-changing security-sensitive action that can alter credentials in the local environment and broaden the agent's effective access, creating a mismatch that may mislead users and downstream policy enforcement.

Description-Behavior Mismatch

Medium

Confidence: 78% confidence
Finding: The documentation says RBAC auditing is unavailable in the free edition, yet later references commands like `az role assignment list` as validation guidance. This inconsistency can confuse operators and agents about the allowed security scope, potentially leading to execution of higher-sensitivity IAM inspection actions under a tool advertised as limited-function.

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The skill first states it lacks security auditing and compliance checks, but later claims 'built-in security checks' and compliance protections in comparison tables. Contradictory security claims are dangerous because users or supervising agents may overtrust the tool's protections, skip independent validation, or authorize broader use based on capabilities the skill does not reliably provide.

Vague Triggers

Medium

Confidence: 74% confidence
Finding: The skill uses broad capability language without tight trigger constraints, which can cause an agent to over-apply the skill to loosely related prompts and execute local CLI commands in situations the user did not clearly intend. In an `exec`-capable skill, ambiguous scope increases the chance of unsafe command execution, especially when combined with session or subscription management features.

Static analysis

No suspicious patterns detected.