xss-html-injection
Warn
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONOBFUSCATIONDYNAMIC_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill provides numerous templates and examples for exfiltrating sensitive browser data to external domains.
- Evidence: Scripts designed to capture
document.cookieand send it tohttps://collector.invalid/stealorhttps://collector.invalid/capture. - Evidence: A keylogger implementation that sends captured keystrokes to
https://collector.invalid/log?k=via Image objects. - [OBFUSCATION]: The skill includes instructions on how to bypass security filters using various encoding and obfuscation methods.
- Evidence: Use of Base64 encoding to hide script content (e.g.,
atob('YWxlcnQoMSk=')). - Evidence: Use of HTML entity encoding (
a), hex encoding (a), and Unicode escapes (\u0061) to obscure payloads. - Evidence: JavaScript string concatenation (
'al'+'ert(1)') used to evade simple string-matching filters. - [DYNAMIC_EXECUTION]: The skill demonstrates techniques to execute dynamically generated or decoded code.
- Evidence: Use of
eval()in combination withatob()to execute Base64-decoded commands. - Evidence: Use of
Function()and[].constructor.constructor()as alternative execution sinks for XSS payloads.
Audit Metadata