The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's Python scripts (xsw_test.py, comment_inject.py, and xxe_inject.py) use subprocess.run to invoke the system curl utility for transmitting manipulated SAML assertions to external endpoints.
[DATA_EXPOSURE]: The xxe_inject.py script constructs an XML External Entity (XXE) payload targeting the sensitive system path file:///etc/passwd. This is intended to test for vulnerabilities on the target service provider.
[REMOTE_CODE_EXECUTION]: The documentation references external security tools and research repositories, specifically SAMLRaider on GitHub and PortSwigger's security academy. These are recognized as well-known security resources.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted SAML assertions via the SAML_RESPONSE_B64 input variable. Ingestion points: Input variable SAML_RESPONSE_B64 in SKILL.md. Boundary markers: None present. Capability inventory: File system access to /tmp, execution of curl via subprocess, and regex manipulation of XML. Sanitization: The skill decodes the assertion but does not perform validation or sanitization, as its purpose is to manipulate the data for testing.

web-vuln-saml-sso