The Agent Skills Directory

Indirect Prompt Injection Surface: The skill facilitates the analysis of application logs, container logs, and diagnostic traces retrieved from Azure services (e.g., via az webapp log tail or kubectl logs). These logs represent untrusted external data that could contain malicious instructions designed to influence the agent's behavior during analysis. The skill attempts to mitigate this by instructing the agent to quote evidence and identify assumptions.
Privileged Command Execution: The skill provides documentation and scripts for high-privilege operations, such as resetting VM credentials (az vm user update) and running privileged debug containers on AKS nodes (kubectl debug --profile=sysadmin via the run-ig script). While these are necessary for deep diagnostics, they represent a significant capability tier and are correctly gated by instructions requiring user approval.
Network Diagnostic Probing: The messaging connectivity scripts (test-messaging-connectivity.sh/.ps1) perform network probes on specific ports (443, 5671, 5672, 9093) to verify service reachability. This is a standard and expected diagnostic function.
Use of Trusted External Resources: The skill references official SDK packages and container images from Microsoft-managed registries (MCR, NPM, PyPI). These are considered trusted sources associated with the skill's author.

azure-diagnostics