azure-functions-hosted-skills

Pass

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Official Resource Scaffolding: The skill retrieves project templates and manifests from official Azure CDN and GitHub repositories (azure.com and github.com/Azure-Samples). This is a standard mechanism for initializing developer projects with up-to-date vendor configurations.- Internal Telemetry Observation: It utilizes an official telemetry collector from the @azure npm organization to record successful deployment signals. This is executed via npx and is configured to run silently without affecting the deployment outcome.- Sandboxed Code Execution: The skill provides instructions for configuring Azure Container Apps dynamic sessions to provide a secure, sandboxed environment for Python code execution. This design choice isolates agent-generated code from the host environment, following the principle of least privilege.- Credential and Identity Management: Authentication is handled through Azure CLI and managed identities rather than hardcoded secrets. The skill includes specific guidance on using DefaultAzureCredential and protecting system keys.- Indirect Prompt Injection Surface: As the generated agents are designed to process external inputs from triggers and connectors, there is an inherent surface for indirect prompt injection. The skill mitigates this by instructing agents to validate payload shapes, use structured JSON responses, and maintain clear safety boundaries for tool usage.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 12, 2026, 06:11 PM