The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file includes shell command placeholders (!find ...) to locate YAML schema and mapping files within the current directory. These commands are restricted to listing local files for agent context and do not process untrusted arguments.
[COMMAND_EXECUTION]: The intended workflow requires the agent to execute bundled Python scripts for data extraction and conversion, as well as standard development tools like pip, git, and infrahubctl.
[EXTERNAL_DOWNLOADS]: The skill fetches vendor device models from the official community-maintained devicetype-library on GitHub.
[INDIRECT_PROMPT_INJECTION]: The conversion process ingests external YAML data from NetBox instances and public libraries.
Ingestion points: Data is read by netbox_to_infrahub_templates.py and the materialize_module_ports.py generator script.
Boundary markers: The skill instructs the agent to analyze coverage reports and explain data loss to the user, providing a verification step.
Capability inventory: Scripts have the ability to read and write local files and communicate with configured NetBox and Infrahub API endpoints.
Sanitization: The transformation logic is governed by structured mapping profiles that explicitly bind NetBox fields to Infrahub attributes, preventing untrusted data from directly influencing command execution or control flow.

infrahub-converting-netbox-device-types