The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent on using Ansible CLI tools such as ansible and ansible-playbook to perform administrative tasks on remote systems. This includes the use of modules like ansible.builtin.shell for arbitrary command execution on managed nodes.
[EXTERNAL_DOWNLOADS]: The documentation includes steps for installing the Ansible software via official package managers including pip, apt, dnf, and brew. It also provides instructions for downloading third-party roles and collections from Ansible Galaxy, the official community registry for Ansible content.
[CREDENTIALS_UNSAFE]: The skill references standard patterns for using SSH private keys (e.g., ~/.ssh/ansible_ed25519) for remote authentication. It suggests the use of ansible-vault for secret management to prevent the storage of plaintext credentials. It also notes that disabling host key verification may be necessary in dynamic environments, which is a common tradeoff in automation.
[PROMPT_INJECTION]: The skill defines an indirect prompt injection surface where external data may be processed. 1. Ingestion points: Data entering via variables in vars, vars_files, and command-line extra variables (-e). 2. Boundary markers: No specific delimiters are suggested to separate variable data from instructions. 3. Capability inventory: The skill enables remote shell execution and configuration changes across a network of hosts. 4. Sanitization: The provided examples do not include explicit validation or sanitization of input variables.

ansible