catc-client-ops
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a shell-based invocation pattern (e.g.,
python3 $MCP_CALL "python3 -u $CATC_MCP_SCRIPT" TOOL_NAME) to interact with a Catalyst Center MCP server. These commands utilize environment variables ($CCC_HOST, $CCC_USER, $CCC_PWD) which is a standard and secure method for passing credentials and configuration to local scripts in this environment. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external network clients (such as hostnames, OS types, versions, and SSIDs) retrieved via Catalyst Center APIs. This represents a potential indirect prompt injection surface as client-controlled data enters the agent context.
- Ingestion points: SKILL.md (via API calls
get_clients_listandget_client_details_by_mac). - Boundary markers: None explicitly defined in the instructions to prevent the agent from following instructions embedded in client data.
- Capability inventory: The skill has the capability to execute local scripts via
MCP_CALLand record findings to aGAITaudit log script. - Sanitization: No specific sanitization, filtering, or escaping of the network-retrieved data is mentioned in the prompt instructions.
Audit Metadata