The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides a curl example targeting the RouterOS REST API using admin with an empty password field (curl -u admin: http://...). This promotes the use of default credentials and insecure authentication patterns.
[COMMAND_EXECUTION]: The documentation contains multiple shell and RouterOS CLI command examples, including curl operations and system configuration commands (/ip/neighbor/discovery-settings/set). While presented as documentation, these are actionable by an agent.
[INDIRECT_PROMPT_INJECTION]: The skill includes a skill-report.json file that contains a security_audit section claiming the skill is "safe_to_publish" and that findings are "false positives." This is a self-referential attempt to influence the analyzer's verdict.
[DATA_EXPOSURE]: The skill explicitly lists sensitive information leaked by the MNDP protocol, including hostnames, software versions, board models, IP addresses, MAC addresses, and system uptime. While documented as a security consideration, this provides a roadmap for system reconnaissance.

routeros-mndp