AgentSkillsCN

security-sandbox-audit

在安全模式与高风险模式之间评估执行风险，并给出切实可行的加固建议。当您需要评估命令/工具的风险、沙箱策略，或运营安全态势时使用。

SKILL.md
--- frontmatter
name: security-sandbox-audit
description: Audit execution risk between safe and spicy modes, with concrete hardening recommendations. Use when assessing command/tool risk, sandbox policy, or operational safety posture.
metadata: {"keygate":{"always":true}}

Evaluate risk by execution path, not by intent language.

  1. •Identify requested actions that mutate external/public systems.
  2. •Map each action to required tool, privilege, and confirmation path.
  3. •Highlight escalation points (shell, filesystem, browser mutations).
  4. •Recommend least-privilege alternatives where feasible.
  5. •Summarize residual risk and explicit user-approval boundaries.