Audit History

Two findings are confirmed because the workflow generates and runs an Azure CLI command from user intent without an explicit validation or confirmation checkpoint. The other 29 detections are Markdown syntax, tool names, flags, or read-only examples, and no prompt injection or malicious intent was found.

All static findings are false positives caused by Markdown inline code, code fences, and visible Azure CLI examples in SKILL.md. No evidence found of prompt injection, credential exfiltration, mutation workflows, or concealed command execution.

The static hits are Markdown inline code, code fences, Azure MCP tool names, and read-only Azure CLI examples in SKILL.md. I found no prompt injection, hidden scripts, credential handling, or data exfiltration intent in the scanned files.

Static analysis flagged many command and crypto-like patterns, but review shows the hits are Markdown examples, KQL snippets, and text matches such as description or desc. No prompt injection, credential exfiltration, mutation workflow, or malicious intent was found. The remaining risk is operational: the skill can guide execution of Azure CLI read-only inventory queries that may expose tenant metadata to the active AI session.

All 95 static findings are false positives. The skill contains documentation and reference content for Azure Resource Graph queries. The detected patterns are: (1) Backtick usage in markdown for code formatting - not actual shell execution; (2) KQL operators like '=~' misidentified as weak crypto - actually case-insensitive comparison; (3) Error handling references misidentified as reconnaissance. This is a legitimate Microsoft Azure documentation skill with no security concerns.