The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains several shell commands that pipe the output of curl requests into python3. While automated scanners flagged these as remote code execution, manual analysis confirms these are safe data-processing patterns. The commands target localhost APIs for Prometheus and Alertmanager to format JSON output using the json.tool module or to iterate through metric labels using small inline Python scripts (-c flag). Because the data source is local and the execution method is for data transformation rather than executing remote scripts, this does not pose a security risk.
[COMMAND_EXECUTION]: The skill includes legitimate administrative commands for managing an OpenStack monitoring stack. These include kolla-ansible for service deployment, docker for container verification and restarts, and firewall-cmd for opening exporter ports. These operations are appropriate for the skill's stated purpose and operate within expected administrative boundaries.
[CREDENTIALS_UNSAFE]: Security best practices are followed for credential management. Examples use Ansible Vault placeholders (e.g., {{ vault_grafana_admin_password }}) or generic descriptive placeholders for third-party integrations like PagerDuty. No hardcoded API keys, passwords, or private tokens are present.
[DATA_EXFILTRATION]: Network activity is confined to querying local monitoring services on the management network (ports 3000, 9090, 9093). There are no attempts to send data to external or untrusted domains.

openstack-monitoring