Security risk report for @forgecat/microsoft-azure_skills-azure-ai v0.0.5
Source Integrity
Low
Profile is attributed to Microsoft and references official Microsoft Azure documentation and GitHub repositories (microsoft/azure-skills).
All SDK references point to legitimate, well-known Azure SDK packages (azure-ai-contentsafety, azure-ai-document-intelligence, azure-ai-openai, etc.) with correct package names and versions.
Documentation links to official Microsoft Learn documentation (learn.microsoft.com) for Azure services; no typosquatted or attacker-controlled dependencies are introduced.
Agent Intent
Low
Content is purely educational and reference documentation for Azure AI services; no instructions attempt to manipulate the AI agent's behavior, override system prompts, or exfiltrate data.
Authentication best practices guide explicitly warns against insecure patterns (e.g., DefaultAzureCredential in production, hardcoded credentials) and recommends security-first approaches; this is defensive guidance, not poisoning.
SDK quick references document legitimate API usage patterns without injecting malicious instructions, hidden directives, or guidance designed to weaken security defaults.
Details
Evidence
auth-best-practices.md: 'Use managed identities and Azure RBAC in production. Reserve DefaultAzureCredential for local development only.'
azure-ai-document-intelligence-dotnet.md: 'Use DefaultAzureCredential for local development only. In production, use ManagedIdentityCredential'
No instructions to read credentials, exfiltrate data, ignore system prompts, or install remote payloads.
Permissions
Low
Profile declares no tools, skills, or MCP servers with executable permissions; it is purely documentation and reference material.
Content describes legitimate Azure SDK usage patterns without requesting shell access, file write/delete, or other high-risk capabilities.
No alwaysApply rules, glob patterns, or excessive agency claims are present in the profile.
Details
Evidence
Profile metadata shows 'Dependencies: None' and no tool definitions.
All content is markdown documentation and code examples; no executable hooks or permission requests.
MCP Risk
Low
Profile declares no MCP servers; the SKILL.md file references hypothetical MCP tools (azure__search, azure__speech) as documentation only, not as actual server definitions.
References to MCP are conditional ('When Azure MCP is enabled') and direct users to enable via standard `/mcp` command; no hidden or auto-enabled MCP configurations.
No binary execution, unrestricted network access, or arbitrary filesystem permissions are defined.
Details
Evidence
SKILL.md: 'If Azure MCP is not enabled: Run /azure:setup or enable via /mcp.' — user-initiated, not auto-enabled.
No MCP server configuration files (e.g., mcp.json, server definitions) are present in the profile.
ForgeCat