pyats-junos-routing
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands using
python3and a platform-provided MCP caller script to run network operations, as shown in the call examples. - [DYNAMIC_EXECUTION]: The skill utilizes shell environment variables (
PYATS_PYTHON,PYATS_MCP_SCRIPT) to dynamically construct the command-line string for execution, which is a common but sensitive pattern for runtime path and interpreter selection. - [INDIRECT_PROMPT_INJECTION]: The skill processes operational data from network devices (e.g., OSPF neighbors, BGP summaries, route tables) which could contain malicious strings injected by external actors into network protocol metadata.
- Ingestion points: Untrusted data enters the agent context via the output of
pyats_run_show_commandas defined inSKILL.md. - Boundary markers: Absent. The skill instructions do not provide specific delimiters or warnings to the agent to disregard potential instructions embedded within the retrieved device output.
- Capability inventory: The skill has the capability to execute system commands (
ping,traceroute) and network-wide show commands via thepyats_run_show_commandtool. - Sanitization: Absent. There is no evidence of data filtering or sanitization applied to the operational output before it is processed by the AI agent.
Audit Metadata