The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's methodology involves the ingestion and processing of untrusted data from external targets, which establishes an attack surface for indirect prompt injection.
Ingestion points: The agent is instructed to capture and analyze SAMLResponse values, JWT tokens, and HTTP headers from external systems (SKILL.md).
Boundary markers: The instructions do not define clear delimiters or provided warnings to isolate the agent from potentially malicious instructions embedded within the analyzed data.
Capability inventory: The skill provides functionality to perform network operations via curl, execute code via python, and use system utilities like grep and subfinder (SKILL.md).
Sanitization: There are no explicit validation or sanitization steps mentioned for processing the external payloads before they are used in script templates.
[DYNAMIC_EXECUTION]: The skill provides Python and Bash script templates meant to be populated with external data and executed locally.
Evidence: The 'SAML signature stripping' and 'JWT none algorithm attack' sections contain Python code segments that utilize standard libraries (base64, re, json) to manipulate external input at runtime.
Risk: The execution of scripts generated from external templates, while intended for security testing, represents a lower-severity dynamic execution risk if the input contains unexpected patterns.

hunt-auth-bypass