The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill acts as a repository of technical attack vectors, providing a capability inventory for potential system exploitation if instructions are interpreted by the agent during processing.
Ingestion points: Instructions and payloads contained within SKILL.md.
Boundary markers: Absent. The skill does not provide clear delimiters or warnings to the agent to treat the included payloads as inert data.
Capability inventory: Provides functional payloads for local file disclosure (targeting paths like /etc/passwd and C:\boot.ini), Server-Side Request Forgery (SSRF), and Denial-of-Service (e.g., Billion Laughs and YAML expansion attacks).
Sanitization: Absent. The skill provides raw, functional injection strings designed to exploit XML parsers.
[COMMAND_EXECUTION]: The skill provides instructions for shell-based payload manipulation using standard utilities.
Evidence: Documentation includes the use of iconv to convert payload encoding (e.g., cat utf8exploit.xml | iconv -f UTF-8 -t UTF-16BE > utf16exploit.xml) to facilitate WAF bypass.
[EXTERNAL_DOWNLOADS]: References several specialized external security tools and repositories for vulnerability research.
Evidence: The skill links to established GitHub repositories including staaldraad/xxeserv, enjoiz/XXEinjector, and BuffaloWill/oxml_xxe.

payloads-xxe