firewall-manager
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to maintain a local audit trail and perform comparisons. It creates state directories and writes JSON snapshots of the firewall configuration using
mkdirandprintfto ensure a rollback path exists.\n- [DYNAMIC_EXECUTION]: A Python utility script is embedded in the instructions to perform structural JSON diffing. This allows the agent to verify that only intended changes were applied to the firewall configuration after a mutation.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from the network controller which could contain malicious instructions in metadata fields.\n - Ingestion points: Firewall policy names and descriptions retrieved from the
unifi-networkMCP server inSKILL.md(Step 5 and Step 7).\n - Boundary markers: Absent; the skill does not wrap policy data in specific delimiters.\n
- Capability inventory: Extensive configuration capabilities including creating, updating, and reordering firewall policies across the network.\n
- Sanitization: Not explicitly performed on the administrative metadata retrieved from the controller.
Audit Metadata