The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes high-privilege network tools and a custom script (scan_private_network) using sudo, allowing it to perform deep network inspection and potentially execute arbitrary code with root privileges.
[CREDENTIALS_UNSAFE]: Explicitly requires sudo permissions for the nmap utility and the custom discovery script, which bypasses standard user-level constraints and may lead to administrative credential exposure.
[COMMAND_EXECUTION]: The skill utilizes sensitive system utilities including nmap, traceroute, dig, curl, and ssh-keyscan with scripts to interact with internal network services, which can be leveraged for unauthorized reconnaissance or service disruption.
[DATA_EXFILTRATION]: The skill is designed to map internal network topology and identify services, versions, and vulnerabilities on private hosts, performing extensive internal reconnaissance that reveals sensitive infrastructure details to the agent context.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the network environment which could influence agent behavior during report generation.
Ingestion points: Host responses from curl (HTTP headers), nmap (service banners), and dig (DNS zone transfers) are ingested in Phase 2.
Boundary markers: The skill does not define clear boundaries or 'ignore' instructions for the data interpolated into the report templates.
Capability inventory: The agent possesses elevated privileges (sudo), file system access (mkdir, cp, rm), and active network scanning capabilities.
Sanitization: No sanitization or escaping of the retrieved network data is performed before it is written to the final markdown report.

Private Network Security Scan