pipelines-security-audit
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data retrieved from Azure DevOps APIs (project names, service connection details, etc.). This content is untrusted and could potentially contain malicious instructions intended to influence the agent's reporting or behavior.
- Ingestion points: Azure DevOps API endpoints at
dev.azure.com(SKILL.md). - Boundary markers: The skill instructs the agent to report findings in specific tables, which provides some structural separation, but no explicit LLM-specific boundary markers are used for the API data.
- Capability inventory: Network access via
curlfor GET requests. - Sanitization: Includes instructions for URL encoding and shell escaping for variable expansion.
- [COMMAND_EXECUTION]: The skill relies on executing
curlcommands in the user's terminal to fetch security metadata. While it targets a well-known service (Microsoft Azure DevOps), it requires the execution of multiple network-bound shell commands. - [DYNAMIC_EXECUTION]: For Windows PowerShell environments, the skill instructs the agent to dynamically create a temporary file (
body.json) containing JSON data to be passed to thecurlcommand. This represents runtime generation of data that influences command execution.
Audit Metadata