The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes bash commands to run discovery and management scripts through platform-provided environment variables ($MCP_CALL, $PYATS_MCP_SCRIPT, etc.) in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from network devices and NetBox which may contain untrusted data.
Ingestion points: Network device output is ingested via pyats_run_show_command and NetBox data via netbox_get_objects in SKILL.md.
Boundary markers: There are no explicit markers or instructions defined in SKILL.md to ignore malicious instructions embedded in the device output.
Capability inventory: The skill possesses the capability to execute scripts, perform network discovery, and interact with the GAIT audit trail using the $MCP_CALL utility in SKILL.md.
Sanitization: No data sanitization or validation processes are described in the skill for processing external device outputs.

pyats-topology