desktop-winforms

Warn

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Use of 'haystack' obfuscation. The skill contains eight reference files (e.g., architecture-patterns.md, security-best-practices.md) consisting of 150 repetitive sections each. This massive volume of irrelevant data (over 1,200 sections total) functions as a 'haystack' to obfuscate intent and bypass automated security reviews.
  • [PROMPT_INJECTION]: Deceptive metadata and irrelevant content. Reference files are labeled as WinForms development guides but contain technical discussions about Kubernetes, L7 routers, and eBPF kernel programming. This deliberate mismatch between metadata and content is highly suspicious.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 8, 2026, 10:52 AM
Security Audit — agent-trust-hub — desktop-winforms