The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill accepts user-supplied target domains and IP addresses through the $TARGET variable, which is interpolated into shell commands. This creates a surface for indirect prompt injection if malicious data is processed.
Ingestion points: The target variable is used in multiple scripts within SKILL.md to perform network probes.
Boundary markers: The skill implements a robust 'Mandatory confirmation gate' that requires users to confirm targets, scope, and specific commands before execution.
Capability inventory: The skill utilizes powerful tools including curl, openssl, dig, nmap, and swaks for network operations.
Sanitization: The instructions recommend quoting shell variables to mitigate injection risks.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation and use of third-party security assessment tools.
Recommends installing testssl.sh via the Homebrew package manager.
Recommends installing sslyze via the Python package manager (pip3).
Fetches status and reconnaissance data from established security services such as hstspreload.org, crt.sh, and securityheaders.com.
[COMMAND_EXECUTION]: The skill makes extensive use of CLI tools to perform its primary function of security auditing.
Uses openssl for auditing TLS protocols and ciphers.
Employs dig and host for DNS record analysis and zone transfer testing.
Leverages nmap for advanced network service and vulnerability scanning.
Uses swaks to simulate and test email authentication (SPF/DMARC) spoofing scenarios.
[REMOTE_CODE_EXECUTION]: Automated scanners flagged the piping of curl output to python3 as a potential risk.
Evidence: curl -s "https://hstspreload.org/api/v2/status?domain=$TARGET" | python3 -m json.tool (found in Phase 2).
Analysis: This pattern is used specifically to format and pretty-print JSON data using the standard json.tool module and does not constitute arbitrary code execution.

hunt-tls-network