claude-on-foundry
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates infrastructure management by instructing the agent to run Azure CLI (
az) and Azure Developer CLI (azd) commands. These include resource provisioning (azd up), environment teardown (azd down), and role assignment management. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user input to configure deployment parameters like legal entity names, regions, and model families. It mitigates potential injection risks by including a 'Safety checklist' that mandates user confirmation for destructive operations (like
az cognitiveservices account purge) and prohibits bypassing pre-provisioning hooks. - [DATA_EXPOSURE_AND_EXFILTRATION]: The skill interacts with local configuration files including
~/.claude/settings.jsonand.vscode/settings.json. It treats these as sensitive, advising the agent to show contents and get user approval before making changes. It also contains explicit warnings against storing API keys or tokens in tracked files. - [EXTERNAL_DOWNLOADS]: The skill references the
Azure-Samples/clauderepository and official Microsoft documentation. These are well-known, trusted sources associated with the skill's authoring organization (Azure Samples).
Audit Metadata