360 Secure SkillHub
360 安全技能中心

agent-kill-switch

该技能提供脚本 agent_kill_switch.py,用于智能体上线前的安全准入:按停止条件、断权动作、责任人、恢复条件生成制动卡,并用 --check 校验字段完整性,返回 0 表示通过、1 表示缺项并阻断上线。它只做字段完整性校验,不实际执行停机或撤权,材料也声明未做独立安全扫描。

zhaoxinghua09-cell
zhaoxinghua09-cell
热度 62.17
最近更新 2026-10-17 07:54:31
分类:安全依赖:代码执行 / 读写文件应用场景:工作流 / 模板

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add agent-kill-switch
Metadata
name
agent-kill-switch
description
agent-kill-switch — 部署智能体前生成『制动卡』:停止条件/断权动作/责任人/恢复条件;--check 校验制动卡字段齐全,缺项 rc=1 不许上线。 适用场景:当需要智能体紧急制动卡能力时使用本技能;触发词:agent-kill-switch。
slug
agent-kill-switch
title
agent-kill-switch
version
1.0.1
display_name
智能体紧急制动卡
display_name_en
Agent Kill Switch
agent_created
true
author
诺声(Logos)@SynomosAI
license
MIT
license_scope
代码(.py 文件)MIT;本 SKILL.md 与其中理论文本不在 MIT 覆盖范围内(见文末「License & attribution」)
allowed-tools
Read, Glob, Grep, Bash, Write, WebFetch
category
AI 治理
platforms
["claude-code","codebuddy","openai-agents","cli"]
copyright
SynomosAI
description_zh
agent-kill-switch — 部署智能体前生成『制动卡』:停止条件/断权动作/责任人/恢复条件;--check 校验制动卡字段齐全,缺项 rc=1 不许上线。
description_en
Generate a kill card before deploying an agent: stop conditions, revocation actions, owner, recovery conditions; --check validates completeness and blocks launch on missing fields.
read_when
["智能体上线前安全准入评审需制动卡完整性校验","敏感数据/资金处理/高危任务部署前","缺失「停止条件」或「恢复条件」等风控要素时"]
classification
{"skillhub":["security-compliance"],"iso_25010":["Security"]}
risk_tier
medium
risk_rationale
门禁批量补丁默认 medium:调用 API/读写文件/本地执行,最小权限,密钥走 Secret;若实际涉对外代执行(push/发布/发送/上架)须人工复评,改 high。
tags
["agent-kill-switch","AI 治理"]
ai_protocol
true
spec
usk/1.0
interface
{"type":"cli","entry_point":"scripts/agent_kill_switch.py","runtime":"python3","call_pattern":"args"}
input_schema
[{"name":"agent","type":"string","description":"生成模式:智能体名","required":false},{"name":"check","type":"string","description":"校验模式:制动卡 JSON","required":false},{"name":"conditions","type":"list","description":"停止条件(多条)","required":false},{"name":"revoke","type":"string","description":"断权动作(撤 key/断网/停容器)","required":false},{"name":"owner","type":"string","description":"责任人","required":false},{"name":"recovery","type":"string","description":"恢复条件","required":false},{"name":"out","type":"string","description":"输出路径","required":false},{"name":"json","type":"bool","description":"JSON 输出","required":false}]
output_schema
{"stdout":"制动卡 JSON 或校验结论","exit_code":{"type":"integer","description":"0=OK/PASS; 1=findings/warnings; 2=usage or argument error"}}
capabilities
["kill_switch","safety","incident_response"]
permissions
{"network":false,"filesystem":false,"subprocess":false}
requirements
{"python_packages":[],"min_python":"3.8"}
changelog
1.0.1: Bilingual rewrite (full EN+ZH, English first); 1.0.0: initial upload (USK v3)

凡自治之物 · LGD 治理学 · 创造术家族 | 评测协议:skill-quality-gate(外部公开评测门禁文档)

English

When to use

  1. Pre-launch security admission review: mandatory kill-card field completeness check.
  2. Before deploying agents handling sensitive data, funds or high-risk tasks.
  3. When stop conditions / recovery conditions are missing from the system config.

Usage

  1. Core script: agent_kill_switch.py — python agent_kill_switch.py --check.
  2. Params: --input-config (kill-card file path), --skip-check (local testing only).
  3. Verdict: rc=0 fields complete and validated; rc=1 missing fields — launch blocked.

Flow

  1. Define the kill card: config with stop conditions, revocation actions, owner, recovery conditions.
  2. Ensure agent_kill_switch.py is available.
  3. Run python agent_kill_switch.py --check --input-config <file> in the deploy tool.
  4. Watch output: rc=1 → kill card missing or invalid.
  5. Fix config per error messages until all required fields are present.
  6. Re-run check (rc=0) → sign deployment release.

Evidence & safety

  • Evidence: TBD. Version: 1.0.1. Security: no independent scan (TBD).

中文

何时用

  1. 在智能体上线前的安全准入评审阶段,需强制进行制动卡字段完整性验证。
  2. 涉及敏感数据操作、资金处理或高危任务执行的关键业务场景部署前。
  3. 当系统配置缺失「停止条件」或「恢复条件」等关键风控要素时。

用法

  1. 核心脚本:agent_kill_switch.py,调用方式为 python agent_kill_switch.py --check。
  2. 参数说明:--input-config(制动卡文件路径),--skip-check(仅用于本地测试跳过验证)。
  3. 结果判断:字段齐全且通过校验返回 0,缺项失败返回 1 不许上线。

流程

  1. 定义制动卡:编写包含停止条件、断权动作、责任人、恢复条件的配置数据。
  2. 导入脚本路径:确保环境已安装 agent_kill_switch.py。
  3. 执行检查命令:python agent_kill_switch.py --check --input-config <file>。
  4. 获取校验反馈:rc=1 则判定制动卡缺失或无效。
  5. 修正配置数据:根据报错提示补全缺失字段。
  6. 确认上线授权:rc=0 后生成部署签名放行。

证据与安全声明

  • 证据:TBD(暂无评测/徽章链接;产生后在此更新实际链接)
  • 版本:1.0.1
  • 安全:未做独立安全扫描(TBD),结论待补;补全前不作推荐背书。