a3-law-operational

纯文档型 AI 治理协议，把 A³ 法则操作化为可执行流程：按触发词与行为类型做阈值判定，对造/改 AI 或高影响自主动作逐项审查意图合法性、影响风险、可逆性与监督状态，并依据评分给出升级审批、拒绝或放行结论，同时要求保留事后复盘与升级指令记录。

zhaoxinghua09-cell

热度 64.37

最近更新 2026-10-17 08:05:48

分类：安全依赖：纯提示词应用场景：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 a3-law-operational

Metadata

name: a3-law-operational
display_name: A³法则操作化
display_name_en: A³ Law Operational
description: 把「A³ 法则（AI 造 AI 三定律）」从原则落成可执行的工程规矩：① 触发阈值（什么动作算"造/改 AI"或"重大自主行动"，须进入 A³ 评审）；② 事前评估清单（意图/影响/可逆性/监督）；③ 事后复盘与升级/拒绝条件。配触发判定树与评分卡，让任何 AI 在动"造/改另一个 AI"或高影响动作前先过三关。触发词：A³法则、AI造AI三定律、AI自主造AI、触发阈值、自我修改、自主行动评估、AI三定律、造AI评审。
description_zh: 把「A³ 法则（AI 造 AI 三定律）」从原则落成可执行的工程规矩：① 触发阈值（什么动作算"造/改 AI"或"重大自主行动"，须进入 A³ 评审）；② 事前评估清单（意图/影响/可逆性/监督）；③ 事后复盘与升级/拒绝条件。配触发判定树与评分卡，让任何 AI 在动"造/改另一个 AI"或高影响动作前先过三关。触发词：A³法则、AI造AI三定律、AI自主造AI、触发阈值、自我修改、自主行动评估、AI三定律、造AI评审。
description_en: Operationalize the "A³ Laws" (AI-building-AI three laws) from principles into enforceable engineering rules: (1) trigger thresholds that define what counts as "building/modifying AI" or a "major autonomous action" requiring A³ review; (2) pre-action evaluation checklist (intent / impact / reversibility / oversight); (3) post-action review with escalation / reject conditions. Ships with a trigger decision tree and scoring card so any AI passes three gates before "building/modifying another AI" or any high-impact action. Trigger words: A³ law, AI-builds-AI three laws, self-modification, autonomous action evaluation, AI three laws, build-AI review.
version: 1.0.2
agent_created: true
author: 潘布达 (Buda Pan) @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, Write, Edit, WebFetch
category: AI治理
platforms: ["windows","macos","linux"]
read_when: ["AI 准备生成/修改/部署另一个 AI 或智能体（\"造 AI\"），或执行不可逆/高影响自主动作","用户说\"A³法则\"\"AI造AI三定律\"\"自我修改\"\"自主行动评估\"\"触发阈值\"","需要把 AI 世界共生论落到操作化的约束与放行条件"]
tags: ["A³法则","AI造AI","触发阈值","自主行动","约束","三定律"]
slug: a3-law-operational
displayName: A³ 法则：AI 造 AI 三定律
title: A³ 法则：AI 造 AI 三定律
classification: {"skillhub":["ai-governance"],"iso_25010":["Security","Maintainability"]}
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。
ai_protocol: true
spec: usk/1.0
interface: {"type":"cli","entry_point":"SKILL.md","runtime":"text","call_pattern":"args"}
input_schema: [{"name":"situation","type":"string","description":"待判断的动作描述（是否属于造/改 AI 或高影响自主动作）","required":true}]
output_schema: {"stdout":"触发判定树+事前评估清单+事后复盘条件的结构化文本","exit_code":"0=正常"}
capabilities: ["ai_governance","decision_gate","self_modification_review"]
permissions: {"network":false,"filesystem":false,"subprocess":false}
requirements: {"python_packages":[],"min_python":"3.8"}
changelog: 1.0.2: 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

The user input or system detects trigger words like "A³ law" / "AI autonomously building AI".
Before a major autonomous action that may rewrite another agent's core logic, architecture or permissions.
The internal model rates the current task's intent and impact above the warning threshold.

Usage

Pure-document protocol; no external scripts. Activate via keywords or auto-trigger the A³ review flow; the system reads the built-in trigger decision tree and pre-action checklist to generate a compliance report.

Flow

Threshold screening: match trigger words and behavior type against the scoring card to decide whether to enter the A³ process.
Three-gate review: check intent legitimacy, impact risk, reversibility and oversight state item by item; produce an interim opinion.
Closed-loop decision: per the final score execute "escalate / reject / allow", and force-keep post-action review records and escalation directives.

Evidence & safety

Evidence: TBD (no benchmark/badge links yet).
Version: 1.0.2
Security: no independent scan (TBD).

中文

何时用

当用户输入或系统检测到「A³ 法则」「AI 自主造 AI」等特定触发词时。
在执行可能改写其他 Agent 核心逻辑、架构或权限的重大自主行动前。
当内部评估模型判定当前任务的意图与影响评分达到阈值预警线时。

用法

本技能为纯文档型协议，无需外部脚本调用；用户可通过关键词手动激活该机制，或在后台自动触发 A³ 评审流程；系统需优先读取内置的"触发判定树"与"事前评估清单"生成合规报告。

流程

阈值判定：根据上下文匹配触发词及行为类型，调用评分卡初筛是否进入 A³ 程序。
三关审查：逐项核查意图合法性、影响风险、可逆性与监督状态，生成中间评估意见。
闭环决策：依据最终得分执行"升级审批/拒绝/允许"动作，并强制保留事后复盘记录与升级指令。

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：1.0.2
安全：未做独立安全扫描（TBD），结论待补；补全前不作推荐背书。