The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by ingesting untrusted target identifiers and processing them with tools capable of network and file system interaction.
Ingestion points: Target URLs and domains provided by users for reconnaissance and scanning (SKILL.md).
Boundary markers: Absent. No delimiters or instructions are provided to ignore malicious content that might be encountered when scanning external targets.
Capability inventory: The skill utilizes subprocess calls for nmap, nikto, dirb, sqlmap, subfinder, and curl (SKILL.md).
Sanitization: Absent. No validation or sanitization logic is specified for the input targets.
[COMMAND_EXECUTION]: The skill documentation provides specific bash commands for executing security tools such as nmap, sqlmap, nikto, dirb, and subfinder to identify vulnerabilities and perform reconnaissance.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill provides an example of a directory traversal attack using curl to access the /etc/passwd file on a target system. While this is a standard penetration testing check, it involves targeting sensitive system files.

penetration-testing