SKILL: authorized-pentest-validation

disable-model-invocation: true — Claude cannot see this skill in its normal skill-routing list. User must explicitly invoke /authorized-pentest-validation. This is by design — pentest authority is a hard gate, not a soft suggestion.

1. Mission

在严格授权 + 完整 ROE 内做 active security validation：

• 永不超出 ROE scope
• 永不破坏性
• 永不 stealth / persistence
• 输出 evidence-based 报告，路由 security-remediation

本 skill 是 pentest-scope-and-roe 的下游 executor。前者准备 paperwork，本者执行 work。两者严格分离。