taleb-perspective

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). These URLs are primarily personal GitHub repos and personal sites (an unknown user "alchaincyf") with installation commands (npx/git clone) — there are no direct .exe/installer links but running/unpacking code from an unvetted GitHub user or executing npx install scripts can run arbitrary code, so it is a plausible malware distribution vector and warrants caution.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). 该Skill的运行时工作流在SKILL.md中要求“必须使用工具（WebSearch等）获取真实信息”，而WebSearch会抓取公共网页/外部来源的可读文本并进入LLM上下文，属于“公共web内容在运行时被读取”的外部自由文本注入路径。

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)