data-science-statistical-analysis

Warn

Audited by Gen Agent Trust Hub on Jul 17, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill package contains eight unreferenced documentation files (e.g., 'references/architecture-patterns.md', 'references/security-best-practices.md') that consist of hundreds of identical sections of technical filler text. These files discuss infrastructure topics unrelated to statistical analysis, such as Kubernetes HPA, eBPF, and kernel system calls like epoll_wait and io_uring. This structure acts as 'context stuffing,' saturating the agent's memory with irrelevant information, which can interfere with the execution of primary tasks and mislead the agent's technical grounding.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 17, 2026, 11:35 PM
Security Audit — agent-trust-hub — data-science-statistical-analysis