cloud-architecture

Warn

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a context window flooding attack using extreme redundancy.
  • Evidence: Eight reference files (e.g., 'architecture-patterns.md', 'code-organization.md', 'security-best-practices.md', 'testing-strategies.md') contain exactly 150 repetitive sections each.
  • Mechanics: Each section contains identical technical jargon regarding kernel system calls (epoll_wait, io_uring) and cache locality. Across all files, this results in approximately 1,200 repetitions of the same paragraphs and code snippets.
  • Security Risk: This high-volume 'bloatware' is designed to exhaust the model's context window. This can push critical system prompts or safety guardrails out of the model's short-term memory, potentially allowing for behavior override or restriction bypass.
  • [PROMPT_INJECTION]: The repetitive technical filler text acts as adversarial noise to disrupt the model's reasoning and instruction-following consistency during complex architectural tasks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 13, 2026, 02:15 PM
Security Audit — agent-trust-hub — cloud-architecture