The Agent Skills Directory

[DATA_EXFILTRATION]: The skill identifies sensitive local file paths such as ~/.ssh/id_rsa, /etc/shadow, and C:\Windows\win.ini as typical targets for Local File Inclusion (LFI) vulnerability testing. These are presented as reference points for auditing target applications.
[COMMAND_EXECUTION]: Lists various shell command injection payloads (e.g., ;whoami, $(whoami), and nslookup attacker.com) that are used to verify and exploit remote command execution vulnerabilities in web applications.
[OBFUSCATION]: Employs a defensive documentation technique by inserting brackets into PHP function names (e.g., syst[e]m, passt[h]ru) to prevent the file from being quarantined by antivirus software on the host machine. This methodology is explicitly explained within the file's content.
[INDIRECT_PROMPT_INJECTION]: The skill involves the analysis and processing of untrusted data received from external web servers during security testing, which presents an attack surface for indirect injection.
Ingestion points: Output from network assessment tools mentioned in SKILL.md (such as curl, sqlmap, and nuclei).
Boundary markers: The instructions do not define specific delimiters for separating tool output from the agent's internal logic.
Capability inventory: The skill utilizes network-capable tools and manual command execution for security auditing.
Sanitization: There are no explicit instructions for filtering or sanitizing the data returned from target web applications before processing.
[EXTERNAL_DOWNLOADS]: Provides references to reputable security industry resources including the OWASP testing guide and PortSwigger Web Security Academy. It also lists well-known security research tools and the HackTricks information repository.

testing-web-applications