The Agent Skills Directory

[PROMPT_INJECTION]: The metadata in SKILL.md identifies the author as 'Microsoft', which misrepresents the actual provider 'kinnekate-hub'. This deceptive labeling can lead to an unearned level of trust from users.
[PROMPT_INJECTION]: The skill processes external Azure configuration data, creating an indirect prompt injection surface. Ingestion points include references/services/functions/automation.md (via az config commands). Boundary markers exist as Markdown headers in references/services/functions/assessment.md, but lack explicit directives to ignore instructions. Capability inventory includes resource deletion and code deployment in references/services/functions/automation.md. No sanitization is performed on ingested data before use in scripts.
[DATA_EXFILTRATION]: The migration logic in references/services/functions/automation.md accesses sensitive Azure application secrets and connection strings. While necessary for its primary migration purpose, this creates a data exposure risk.
[COMMAND_EXECUTION]: Automation scripts in references/services/functions/automation.md use shell variables like $appName directly in Azure CLI commands, which could be exploited for command injection if these variables are populated with malicious strings from the environment.
[EXTERNAL_DOWNLOADS]: Fetches the 'resource-graph' extension from Microsoft's official registry and downloads application packages from Azure storage as part of the standard migration workflow.

azure-upgrade