The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill explicitly instructs the agent and user to use sudo for various operations, such as SYN scans (-sS) and OS detection (-O). This involves running a network-facing tool with administrative rights, which increases the potential impact of any vulnerabilities in the tool or misconfigurations in the scan.
[COMMAND_EXECUTION]: The skill provides numerous complex shell commands involving nmap, awk, and shell redirection. While intended for reconnaissance, these commands represent a capability for high-risk system-level execution. If the agent accepts unvalidated hostnames or IP ranges from a user, it could be leveraged to scan internal or unauthorized networks.
[INDIRECT_PROMPT_INJECTION]: The skill provides a Python script to parse Nmap's XML output. This establishes a vulnerability surface where the agent processes data retrieved from external network services. A malicious host could provide crafted service banners or metadata that, when parsed and included in the agent's context, could attempt to influence the agent's subsequent logic.
Ingestion points: scans/lab-2026-10-20.xml is parsed using xml.etree.ElementTree in SKILL.md.
Boundary markers: There are no boundary markers or instructions to ignore embedded content within the parsing script.
Capability inventory: The skill utilizes nmap, sudo, mkdir, and file system writes (-oA, >).
Sanitization: The provided Python example does not include any sanitization or schema validation for the XML content before printing the service names and versions.

nmap-recon