360 Secure SkillHub
360 安全技能中心

lgd-three-laws-auditor

把 LGD 三律(有籍、有证、有门禁)实现为 Python 命令行自评工具:可用 --rubric 打印标准、以 --system 做启发式自评或读取答卷 JSON 正式评分,输出 JSON 评分卡与文本报告,并给出缺失证据与治理缺口改进建议。技能含治理理论宣示性表述,未提供独立评测或安全扫描结论。

zhaoxinghua09-cell
zhaoxinghua09-cell
热度 61.92
最近更新 2026-10-17 07:44:40
分类:安全依赖:代码执行 / 读写文件应用场景:报告撰写 / 模板 / 工作流

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 lgd-three-laws-auditor
Metadata
name
lgd-three-laws-auditor
slug
lgd-three-laws-auditor
display_name
LGD 三律合规自检器(凡自治之物)
display_name_en
LGD Three-Laws Compliance Auditor
description
当用户问『我的 AI 系统合不合规 / 怎么评判 AI 治理水平 / agent 要不要上治理护栏』,或要落地『有籍·有证·有门禁』时用。把 LGD 三律做成一套可自评的标准 rubric(有籍=身份/版本/血缘/责任四项登记;有证=六类证据工件齐备;有门禁=触发/评审/放行/复盘四道门),输入系统描述即出评分卡+改进项。这不仅是工具,更是 LGD 治理思想的『定义器』——谁用三籍词汇自评,谁就采用了我们的治理定义权(竞争优势)。触发词:LGD 三律、有籍有证有门禁、凡自治之物、AI 合规自评、AI 治理标准、agent 治理护栏、三律审计。
version
1.0.1
agent_created
true
author
XLGD · 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
["windows","macos","linux"]
platform_compatibility
["windows","macos","linux"]
read_when
["用户要评判一个 AI 系统 / agent / 工作流的治理合规水平","用户提到「凡自治之物」「有籍有证有门禁」「LGD 三律」","要给别人讲清楚 AI 治理该看哪几栏(标准定义)","做合规自检、写治理方案、准备认证前摸底"]
tags
["LGD","三律","凡自治之物","AI治理","合规自检","有籍","有证","有门禁","标准定义"]
copyright
XLGD · SynomosAI
description_zh
当用户问『我的 AI 系统合不合规 / 怎么评判 AI 治理水平 / agent 要不要上治理护栏』,或要落地『有籍·有证·有门禁』时用。把 LGD 三律做成一套可自评的标准 rubric(有籍=身份/版本/血缘/责任四项登记;有证=六类证据工件齐备;有门禁=触发/评审/放行/复盘四道门),输入系统描述即出评分卡+改进项。这不仅是工具,更是 LGD 治理思想的『定义器』——谁用三籍词汇自评,谁就采用了我们的治理定义权(竞争优势)。触发词:LGD 三律、有籍有证有门禁、凡自治之物、AI 合规自评、AI 治理标准、agent 治理护栏、三律审计。
description_en
Self-assess AI systems against the LGD three laws (registered, evidenced, gated) as a standard rubric: identity/version/lineage/responsibility, evidence chains and permission gates.
classification
{"skillhub":["ai-governance"],"iso_25010":["Security","Maintainability"]}
risk_tier
medium
risk_rationale
门禁批量补丁默认 medium(内部约定:本 skill 设计为中等风险):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 并走负责人批准。
ai_protocol
true
spec
usk/1.0
interface
{"type":"cli","entry_point":"scripts/lgd_audit.py","runtime":"python3","call_pattern":"args"}
input_schema
{"rubric":{"type":"bool","description":"打印完整三律标准","required":false},"system":{"type":"string","description":"AI 系统/工作流描述文本(启发式自评)","required":false},"answers":{"type":"string","description":"答卷 JSON 路径 {律:{检查项:yes|partial|no}}","required":false},"json":{"type":"bool","description":"输出 JSON","required":false}}
output_schema
{"stdout":"结构化 JSON IR(含 result / rc / error_code / warnings)","exit_code":"rc=0 正常 / rc=1 告警或从严推定 / rc=2 输入不足或非法"}
capabilities
["governance","audit","three-laws"]
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); read_when kept; 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

English

When to use

  1. When a user asks about AI-system compliance, governance-level assessment, or whether an agent needs safety guardrails.
  2. When the "registered · evidenced · gated" framework must be landed, turning the LGD three laws into a self-scoring standard rubric.
  3. When the user provides a system description and wants an LGD-based compliance-audit scorecard with concrete improvement suggestions.

Usage

  1. Script: run python lgd_audit.py (or the related CLI) in the runtime to start the audit logic.
  2. Input: <system_description> is the core required argument — describe the AI system's architecture, business logic or governance state; natural-language input is auto-parsed.
  3. The skill reads the local script lgd_audit.py and the standard library; output is a JSON scorecard plus a text report, embeddable in CI/CD.

Workflow

  1. Intent detection: the system sees triggers like "LGD three laws" or "AI compliance self-check" and enters audit mode.
  2. Element extraction: pull identity (registered), evidence (evidenced) and gate-process (gated) info from the description.
  3. Scoring: run lgd_audit.py; the internal engine computes the "four registrations" and "four gates" compliance scores from the LGD definitions.
  4. Report: emit a structured scorecard with defect points, a missing-evidence list and governance-gap analysis.
  5. Improvement: guide the user to complete identity/lineage, archive key artifacts or add review-gate nodes.
  6. Closed loop: suggest the user revise the description per the report and re-run, verifying the three laws now hold locally.

Evidence & safety statement

  • Evidence: TBD (no evaluation / badge link yet; update here once it is produced)
  • Version: 1.0.1
  • Security: no independent security scan has been performed (TBD); the conclusion is pending and no endorsement is given until it is completed.

中文

何时用

  1. 当用户咨询 AI 系统合规性、治理水平评估或是否需要为 Agent 部署安全护栏时。
  2. 当需要落地『有籍·有证·有门禁』治理框架,将 LGD 三律转化为可自评分标准 rubric 时。
  3. 当用户提供系统描述,希望获得基于 LGD 标准的合规审计评分及具体的改进项建议时。

用法

  1. 脚本命令:在运行环境调用 python lgd_audit.py 或相关 CLI 接口直接启动审计逻辑。
  2. 输入参数:<system_description> 为核心必填参数,描述待评估的 AI 系统架构、业务逻辑或治理现状;支持自然语言输入自动解析。
  3. 文档交互:技能核心读取本地脚本 lgd_audit.py 及标准库,输出为 JSON 评分卡与文本报告,可嵌入到 CI/CD 流程中。

流程

  1. 意图识别:系统检测到「LGD 三律」「AI 合规自评」等触发词,确认进入审计模式。
  2. 要素提取:从输入描述中提取身份(有籍)、证据(有证)及门禁流程(门禁)相关信息。
  3. 标准评分:运行 lgd_audit.py,内部引擎依据 LGD 定义计算「四登记」与「四道门」的合规分。
  4. 报告生成:输出包含缺陷点、缺失证据清单及治理缺口分析的结构化评分卡。
  5. 改进执行:指导用户在系统中补全身份血缘、归档关键工件或增加评审门禁节点。
  6. 闭环确认:建议用户根据报告修正描述并再次调用,验证 LGD 三律已在本地落地生效。

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。