csp-bypass-advanced

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The content is educational in nature, providing a matrix of CSP directives and documented bypass techniques for security professionals and researchers.
  • [PROMPT_INJECTION]: The skill contains 'AI LOAD INSTRUCTION' which directs the agent to focus on specific CSP gaps. This is a functional instruction for specialized AI skills and does not attempt to bypass core safety guardrails.
  • [DATA_EXFILTRATION]: While the skill describes theoretical exfiltration channels like DNS prefetch and WebRTC for testing purposes, no logic for exfiltrating user data or hardcoded credentials was found.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 01:23 PM
Security Audit — agent-trust-hub — csp-bypass-advanced