The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses administrative commands like docker exec, docker restart, and kolla-ansible to manage the OpenStack dashboard environment. These actions are standard for the intended user (a cloud operator) and the stated purpose of the skill.
[REMOTE_CODE_EXECUTION]: A troubleshooting step uses curl to fetch data from a local internal service (http://keystone-internal:5000/v3/) and pipes it to python -m json.tool. While piped execution is often flagged by automated scanners, this specific command uses a safe standard library module for JSON formatting rather than executing code, and targets a trusted internal endpoint.
[CREDENTIALS_UNSAFE]: The documentation references paths to sensitive files, including /etc/kolla/admin-openrc.sh and /etc/kolla/passwords.yml. These are the standard locations for OpenStack administrative credentials, and the skill describes their use for legitimate authentication purposes.
[PROMPT_INJECTION]: The skill includes instructions for monitoring and analyzing application and access logs, which represents an indirect prompt injection surface as logs may contain untrusted data.
Ingestion points: docker logs horizon and docker exec horizon cat /var/log/kolla/horizon/horizon.log.
Boundary markers: None present.
Capability inventory: docker exec, docker restart, curl, cat, and kolla-ansible.
Sanitization: None present.

openstack-horizon