The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates and instructions for executing system commands through Ansible, specifically targeting cluster management tools like kubectl and helm.
Evidence: Guidance on using ansible.builtin.command and become: true for administrative tasks in SKILL.md.
Mitigation: The skill explicitly requires setting requires_approval: true for any task using become, changing host/cluster state, or installing packages.
[PROMPT_INJECTION]: As the skill is designed to review, edit, and explain external files (Ansible playbooks), it possesses an attack surface for indirect prompt injection where malicious instructions embedded in a playbook could attempt to influence agent behavior.
Ingestion points: Playbooks located in the ansible/playbooks directory as specified in the SKILL.md frontmatter.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" blocks for the ingested playbook content.
Capability inventory: The skill utilizes subprocess calls via Ansible, privilege escalation (become), and file-write operations.
Sanitization: No explicit sanitization or validation of the ingested playbook content is performed beyond the internal logic of the agent.

playbook-configuration