adf-validation-rules
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Azure Data Factory (ADF) pipeline and dataset JSON files as part of its core validation functionality.
- Ingestion points: The agent reads and analyzes pipeline and dataset JSON files from the repository via the
-PipelinePathand-DatasetPatharguments. - Boundary markers: There are no instructions to use delimiters or ignore embedded natural language instructions within the JSON data being processed.
- Capability inventory: The skill utilizes the agent's ability to execute shell commands (
pwsh) to run thevalidate-adf-pipelines.ps1script. - Sanitization: The instructions do not mention sanitizing or validating the contents of the JSON files to prevent malicious prompts from influencing the agent's behavior during analysis.
- [COMMAND_EXECUTION]: The skill explicitly instructs the agent to execute shell commands using
pwshto run a local PowerShell script (validate-adf-pipelines.ps1). While this is the intended functionality for pipeline validation, it represents a capability that could be targeted if malicious input is provided via the command arguments. - [EXTERNAL_DOWNLOADS]: The skill provides links to official Microsoft Learn documentation for Azure Data Factory components, limits, and best practices.
Audit Metadata