web-xss
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes a methodology for injecting probes and analyzing data reflected from external web applications. This workflow requires the agent to process untrusted data from target application responses, creating a surface for indirect prompt injection.
- Ingestion points: Reflected parameters, HTTP headers, and DOM sinks in target applications as described in SKILL.md.
- Boundary markers: The instructions do not specify delimiters or warnings to ensure the agent ignores embedded instructions within the data it analyzes.
- Capability inventory: The skill metadata references tools such as curl, which provides the agent with network interaction capabilities to execute the described testing methodology.
- Sanitization: No sanitization or escaping of reflected content is mentioned before the agent processes it for analysis.
- [DATA_EXFILTRATION]: The documentation and detection rules contain examples of payloads designed to exfiltrate sensitive data (e.g., document.cookie) to external domains. These are provided as proofs-of-concept for security research and are not executed by the skill scripts.
Audit Metadata