凡自治之物 · LGD 治理学 · 创造术家族 ｜ 评测协议：见平台质量门禁规范
English
When to use
- Before an agent accesses external sensitive tools or performs high-risk operations, when permission boundaries and four-gate rules must be drawn.
- When a project involves code changes, data export or file operations, for automatic anti-overreach and accidental-deletion/accidental-sending checks.
- When potential security-risk patterns or runaway-loop signs are detected, to quickly generate a blocking policy config.
- During the initial setup of new permission tiers in an environment (e.g. isolation, auditing, approval).
Usage
Run python gate_policy.py to invoke the skill's generation logic; it reads gate_policy.py directly. Key parameters: --scenario (business scenario), --tool-list (callable / forbidden tool set), --review-mode (review level). The system auto-detects context from the __pycache__ cache environment and outputs a standard policy-as-code YAML file for configuration.
Workflow
- Analyze permission needs; pin down the agent's callable tool range and the off-limits zones.
- Load
gate_policy.pyand pass in the business-scenario parameters to pre-draft the policy. - Generate the four gate rules (anti-overreach, anti-accidental-deletion, anti-runaway, anti-violation).
- Run internal safety checks to confirm the policy introduces no logic loops or permission conflicts.
- Emit the final
LGD-III_policy.yaml. - Deploy the policy to the runtime environment and verify the permissions take effect.
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
- 当 Agent 计划访问外部敏感工具或执行高危操作前，需划定权限边界与四道门禁规则；
- 在项目涉及代码变更、数据导出或文件操作时，用于自动防越权与误删/误发校验；
- 检测到潜在的安全风险模式或失控循环迹象时，快速生成阻断性政策配置；
- 在环境部署新权限层级（如隔离、审计、审批）的初始化设置阶段。
用法
通过执行 python gate_policy.py 命令调用技能生成逻辑，直接读取文件 gate_policy.py；
关键参数包括：--scenario (定义业务场景)、--tool-list (可调用/禁止工具集合)、--review-mode (评审级别);
系统将基于 __pycache__ 缓存环境自动识别上下文，输出标准 policy-as-code YAML 格式文件供配置使用。
流程
- 分析权限需求，明确 Agent 可调用的工具范围及不可触碰的禁区；
- 加载
gate_policy.py脚本并传入业务场景参数进行政策预演； - 生成四道门禁规则（防越权、防误删、防失控、防违规）；
- 执行内部安全校验，确认策略未引发逻辑死循环或权限冲突；
- 输出最终生成的
LGD-III_policy.yaml文件； - 将策略部署到运行环境并完成权限生效验证。
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。