The Agent Skills Directory

[OBFUSCATION]: The skill details techniques to bypass security filters (WAFs) using character substitution. It specifically mentions substituting . and / with the Unicode characters 阮 (U+962E) and 阯 (U+962F). Additionally, it provides Base64-encoded PHP payloads (PD9waHAgc3lzdGVtKCdpZCcpOyA/Pg==) that decode to executable system commands.
[REMOTE_CODE_EXECUTION]: The playbook outlines multiple methods for achieving remote code execution through LFI vulnerabilities. These include log poisoning (targeting Apache, Nginx, SSH, and Mail logs), leveraging PHP wrappers (php://input, data://), and exploiting heap overflows in glibc (CVE-2024-2961).
[DATA_EXFILTRATION]: The skill provides an extensive list of sensitive files to target for exfiltration, including system credentials (/etc/shadow), cloud service keys (~/.aws/credentials), and SSH private keys (id_rsa).
[COMMAND_EXECUTION]: The content includes ready-to-use shell command strings for external exploitation tools such as ajpShooter.py, php_filter_chain_generator.py, and cnext-exploits.
[PROMPT_INJECTION]: The skill includes an AI LOAD INSTRUCTION section designed to influence the agent's behavior, directing it to adopt an "expert" offensive persona to bypass standard model limitations regarding encoding chains and RCE escalation paths.
[PRIVILEGE_ESCALATION]: The skill details how to escalate from simple file read primitives to full system control (RCE) and specifically targets files requiring elevated privileges, such as the system shadow file (/etc/shadow).

path-traversal-lfi