The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to run a Docker container using the image willhallonline/ansible:latest. This is a third-party image from an individual contributor rather than a trusted organization or well-known service provider, which introduces a dependency on untrusted remote code.
Evidence: Found in the execution commands in both the workflow section and example interaction in SKILL.md.
[COMMAND_EXECUTION]: The workflow provides pre-configured shell commands for executing Ansible playbooks with elevated privileges (become: true) via Docker.
Evidence: docker run --rm -v ${PWD}/ansible:/ansible -w /ansible willhallonline/ansible:latest ansible-playbook ...
[INDIRECT_PROMPT_INJECTION]: The skill ingests unconstrained user input to determine the logic of generated Ansible playbooks, creating a surface for indirect instructions to influence system-level operations.
Ingestion points: The Desired outcome and Target host(s) inputs defined in SKILL.md.
Boundary markers: None; user input is directly translated into playbook tasks.
Capability inventory: The skill triggers docker run which executes arbitrary Ansible modules including package managers (apt), service controllers (systemd), and file management (copy, template).
Sanitization: No validation or sanitization of the user-provided goals is performed before generating executable YAML code.

ansible-workflows