The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains command-line templates for network and web application security assessments.
Evidence: nmap -sV --script vuln TARGET and nikto -h http://TARGET are suggested as preliminary scan tools.
[INDIRECT_PROMPT_INJECTION]: The use of placeholders like TARGET in command examples establishes a potential surface for injection if an agent executes these commands using unsanitized user inputs.
Ingestion points: User-provided values for IP addresses or domain names within the TARGET placeholder (SKILL.md).
Boundary markers: None present in the command templates.
Capability inventory: The skill references nmap, nikto, tcpdump, tail, and systemctl (SKILL.md).
Sanitization: No explicit instructions for sanitizing or validating the TARGET input are provided.
[EXTERNAL_DOWNLOADS]: The skill provides links to penetration testing policies and guidelines from well-known cloud service providers.
Evidence: Includes references to official security documentation from Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP).
[PRIVILEGE_ESCALATION]: Instructional examples include administrative commands that require elevated permissions for system monitoring configuration.
Evidence: Commands such as sudo systemctl restart rsyslog are provided to ensure comprehensive logging is active.
[DATA_EXFILTRATION]: The skill refers to monitoring sensitive system logs as part of the security assessment process.
Evidence: Instructions mention tracking authentication events by monitoring /var/log/auth.log.

Pentest Checklist