The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill involves the agent ingesting and analyzing potentially untrusted XML data from external systems, creating a surface for indirect prompt injection.
Ingestion points: Responses from target XML endpoints, SOAP services, and file uploads described in the methodology (SKILL.md).
Boundary markers: There are no instructions or delimiters provided to the agent to distinguish between application data and potential instructions embedded in the target's XML responses.
Capability inventory: The skill utilizes command execution for security tooling (wfuzz) and network requests to external endpoints.
Sanitization: The instructions do not define any sanitization, validation, or escaping procedures for handling data received from external target systems.
[COMMAND_EXECUTION]: The skill provides explicit examples and instructions for using the wfuzz command-line tool to automate vulnerability scanning against target endpoints.

xxe