The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize large volumes of data from an external Cisco APIC environment, including node status, tenant policies, and fault logs. If an attacker can manipulate descriptions or metadata within the ACI fabric, they could potentially influence the agent's summary or report.\n
Ingestion points: The skill retrieves data via multiple tools: fabric_nodes, tenants_get, fvCtx_get, fvBD_get, fvAp_get, fvAEPg_get, contracts_get, faults, and health in SKILL.md.\n
Boundary markers: The instructions do not define clear delimiters or "ignore instructions" warnings when interpolating ACI data into the audit report.\n
Capability inventory: The skill utilizes shell command execution via python3 $MCP_CALL to run ACI audit tools and visualization scripts (SKILL.md).\n
Sanitization: No sanitization logic for external data is evident in the prompt instructions.\n- [EXTERNAL_DOWNLOADS]: The skill invokes npx -y @drawio/mcp within the Draw.io Fabric Topology section. This command downloads and executes the @drawio/mcp tool from the public npm registry for fabric topology visualization.\n- [COMMAND_EXECUTION]: The skill relies on shell command execution to perform its primary functions, passing sensitive credentials (ACI_USERNAME, ACI_PASSWORD) through environment variables in the command line.

aci-fabric-audit