训

训练数据版权护栏(Data Rights Guard)

作者:注册老炮@MedXpert I SynomosAIv5

当用户说『爬的数据能拿来训练吗』『数据集没写许可证』『微调数据版权合规吗』『怎么确认数据能商用』,或在用数据(爬取/购买/公开集/用户授权)做训练/微调/评测前想确认版权与许可边界时使用。把每条数据当『带权属的资产』:查许可证·商用权限·署名要求·来源可溯,缺许可的不许进训练集。理论根基:LGD 三律(有籍·有证·有门禁)。触发词:数据版权、训练数据合规、数据集许可证、版权护栏、数据权属、商用权限、data license、微调合规。

下载量
289
点赞
73
价格
免费

技能文档

---
name: data-rights-guard
display_name: 训练数据版权护栏(Data Rights Guard)
display_name_en: "Data Rights Guard"
description: "当用户说『爬的数据能拿来训练吗』『数据集没写许可证』『微调数据版权合规吗』『怎么确认数据能商用』,或在用数据(爬取/购买/公开集/用户授权)做训练/微调/评测前想确认版权与许可边界时使用。把每条数据当『带权属的资产』:查许可证·商用权限·署名要求·来源可溯,缺许可的不许进训练集。理论根基:LGD 三律(有籍·有证·有门禁)。触发词:数据版权、训练数据合规、数据集许可证、版权护栏、数据权属、商用权限、data license、微调合规。"
version: 1.0.5
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, Write
category: AI合规
platforms: [windows, macos, linux]
read_when:
  - 要用爬取/购买/公开/用户数据做训练或微调
  - 数据集没写许可证或许可不明
  - 问「这数据能商用吗」
  - 想确认数据来源可溯、署名到位
tags: [数据版权, 训练数据合规, 数据集许可证, 权属护栏, 商用权限, data license, AI合规, 去敏]
slug: data-rights-guard
title: 训练数据版权护栏(Data Rights Guard)
copyright: SynomosAI
description_zh: "当用户说『爬的数据能拿来训练吗』『数据集没写许可证』『微调数据版权合规吗』『怎么确认数据能商用』,或在用数据(爬取/购买/公开集/用户授权)做训练/微调/评测前想确认版权与许可边界时使用。把每条数据当『带权属的资产』:查许可证·商用权限·署名要求·来源可溯,缺许可的不许进训练集。理论根基:LGD 三律(有籍·有证·有门禁)。触发词:数据版权、训练数据合规、数据集许可证、版权护栏、数据权属、商用权限、data license、微调合规。"
description_en: "Copyright and license guard for training data: verify provenance, license terms and commercial-use boundaries before training, fine-tuning or evaluating on scraped, purchased, public or user-authorized data."
classification:
  skillhub: ["security-compliance"]
  iso_25010: ["Security"]
risk_tier: medium
risk_rationale: "门禁批量补丁默认 medium(按默认策略设计为中等风险):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 并走人工批准。"
ai_protocol: true
spec: usk/1.0
interface:
  type: cli
  entry_point: scripts/rights_guard.py
  runtime: python3
  call_pattern: args
input_schema:
  args:
    type: string
    description: "命令行参数透传给主脚本,具体取值见各包 --help"
    required: false
output_schema:
  stdout:
    type: string
    description: "标准输出为 JSON 结果,含逐项检查与总体结论"
  exit_code:
    type: integer
    description: "0=PASS/完成;1=FAIL/有警告或待补强;2=用法或参数错误"
capabilities:
  - cli_tool
permissions:
  network: false
  filesystem: false
  subprocess: false
requirements:
  python_packages: []
  min_python: "3.8"
changelog: "1.0.5: 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:见平台质量门禁规范

## English

### When to use
This skill applies whenever the user needs to confirm copyright and licensing boundaries *before* building a training set, a fine-tuning set or an evaluation set. It is triggered by questions such as "Can scraped data be used for training?", "The dataset states no license" or "How do I confirm the data may be used commercially" — in other words, any question touching data ownership. It also covers large-scale data collection and third-party licensed data: verify that the source is traceable and that authorization credentials are valid against the LGD Three Laws (registered · evidenced · gated), so that data without a license never enters the training set.

### Usage
Two invocation modes are supported: natural-language keyword trigger and direct CLI call. In natural language, a query containing triggers such as "数据版权", "training compliance" or "data license" is automatically parsed and routed to the core logic implemented in `rights_guard.py`. For direct invocation, tune parameters on the command line: `python rights_guard.py --query <question> --source_type [scraped/purchased/public] --permission [commercial/attribution]`. The script lives in `rights_guard.py`; cache files sit under `__pycache__`. Make sure the skill package is loaded in the runtime and that the process holds data-metadata parser permissions.

### Workflow
Execution follows the standard three-laws gate. First, receive the user input and identify the target data source or dataset identifier. Second, read the data metadata / license files and check that the "registered" (source provenance) and "evidenced" (license credentials) fields are complete. Third, compare against the concrete business scenario (e.g. fine-tuning, commercial use) and check for "gated" restrictions and for missing-license risk. Finally, produce a compliance verdict: whether the data may be admitted to the training set, the required attribution format, and a summary of the legal basis. If no valid authorization is detected, the data intake is blocked automatically and the risk is flagged.

### Evidence & safety statement
- Evidence: TBD (no evaluation / badge link yet; update here once it is produced)
- Version: 1.0.5
- 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
本技能适用于用户在构建训练集、微调集或评测集前,需确认数据版权与许可边界的合规场景。当用户提出“爬的数据能拿来训练吗”、“数据集没写许可证”或“怎么确认数据能商用”等涉及数据权属的疑问时,应调用此技能进行审查。此外,在进行大规模数据采集或使用第三方授权数据时,若需要依据 LGD 三律(有籍·有证·有门禁)验证数据来源可溯性与授权凭证的有效性,防止缺许可的数据进入训练集,也属于该技能的核心适用范围。

### 用法
本技能支持自然语言关键词触发与命令行脚本调用两种模式。在自然语言交互中,用户只需输入包含“数据版权”、“训练合规”或 "data license" 等触发的查询,系统会自动解析意图并执行 `rights_guard.py` 中的核心逻辑;在直接调用模式下,可通过命令 `python rights_guard.py --query <问题描述> --source_type [爬取/购买/公开集] --permission [商用/署名]` 进行指定参数调优。脚本位于 `rights_guard.py` 中,相关缓存文件位于 `__pycache__`,需确保运行环境已加载该技能包且具备数据元数据解析器权限。

### 流程
技能执行遵循标准的三律校验流程:首先接收用户输入并识别目标数据来源或数据集标识;其次依据 LGD 理论根基读取数据元数据或许可证文件,核对“有籍”与“有证”信息是否完备;随后对比具体业务场景(如微调、商用),检查是否存在“有门禁”限制及缺许可风险;最后生成合规判定报告,明确是否允许纳入训练集、所需署名格式及法律依据摘要。若检测出无有效授权,则自动阻断数据入库操作并提示风险。

### 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:1.0.5
- 安全:未做独立安全扫描(TBD),结论待补;补全前不作推荐背书。

使用说明

# data-rights-guard · 训练数据版权护栏 / Data Rights Guard

[![LGD Powered](lgd-powered.png)](https://github.com/zhaoxinghua09-cell/lgd-theory)

## English

Every training datum is an *asset with ownership*. Check license · commercial right · attribution · traceable source; entries without permission don't enter the training set. Ships a zero-dependency `rights_guard.py`.

### Why it beats "scrape and train"

| Pain | This skill |
|---|---|
| Scraped data, no license | No-license → excluded |
| Non-Commercial used commercially | NC flag → blocked from commercial |
| Attribution required, not given | Flagged, completed before use |
| "Where did this come from?" | Source traced or quarantined |

### What's inside

- `SKILL.md` — 4-signal ownership model + 3-step gate + iron rule
- `scripts/rights_guard.py` — zero-dependency CLI: scan a manifest.json (license/source/commercial/attribution), rate trainable + advice

```bash
python scripts/rights_guard.py --manifest manifest.json
python scripts/rights_guard.py --manifest manifest.json --json
```

### Benchmark vs alternatives

| Tool | Gap this fills |
|---|---|
| "Just use the dataset" | Per-entry license/commercial gate |
| Manual license reading | Repeatable lexical check |
| Post-infringement cleanup | Pre-training exclusion, not after |

Theoretical root: **LGD Three Laws** (registered · evidenced · gated). Pairs with `doc-desens-scanner` and `ai-policy-radar`.

## 中文

把每条数据当带权属的资产:查许可证·商用权·署名·来源,缺许可的不进训练集。零依赖 `rights_guard.py`。

### 为什么比"抓来就训"强

| 痛点 | 本技能 |
|---|---|
| 爬来的数据,没写许可证 | 无许可证 → 排除 |
| 非商用数据被商用使用 | NC 标记 → 阻断商用 |
| 要求署名却没署名 | 标记,补全后方可使用 |
| "这数据哪来的?" | 来源可溯或隔离 |

### 包内有什么

- `SKILL.md` — 四信号权属模型 + 三步门禁 + 铁律
- `scripts/rights_guard.py` — 零依赖 CLI:扫描 manifest.json(license/source/commercial/attribution),评级可训练并给出建议

```bash
python scripts/rights_guard.py --manifest manifest.json
python scripts/rights_guard.py --manifest manifest.json --json
```

### 与同类工具对比

| 工具 | 本技能填补的空白 |
|---|---|
| "数据集直接用" | 逐条许可证/商用门禁 |
| 人工读许可证 | 可重复的词法检查 |
| 侵权后清理 | 训练前排除,而非事后 |

理论根基:**LGD 三律**(有籍·有证·有门禁)。与 `doc-desens-scanner`、`ai-policy-radar` 配对使用。

---

© 2026 SynomosAI · China · 代码 MIT · 理论文本保留所有权利

---

## License & attribution
**分层许可(务必并读;代码许可不覆盖文档与理论文本)**

- **代码**(本包 `.py` 文件):**MIT**
- **本 `README.md` / `SKILL.md` 及其中的理论文本、方法论与一切理论表述**:**不在 MIT 覆盖范围内**,保留所有权利(All rights reserved)

```
© 2026 SynomosAI · China. All rights reserved.
理论署名 (attribution) : LGD(Lifecycle Governance Doctrine / 全程治理论)— SynomosAI initiative
名称状态 (name status)  : "SynomosAI" — 未申请实体注册、未申请商标注册
                        (not a registered legal entity; no trademark registered)
生产参考部署 (production reference, self-reported) : 无(未声明外部生产参考部署)
                    ← 非认证、非背书、非监管认可(not a certification or endorsement)
代码许可 (code license) : 本包未附 LICENSE 文件(许可待定)
                    本文本与理论表述不在任何代码许可覆盖范围内
引用格式 (cite as)      : 本资产无 DOI
```
(上列为《LGD 对外表述规范》§3.2 统一块,**整体复制、未删改**;发布/重提前须照最新版本复核。)

如何安装此技能?

访问技能市场,点击「安装」按钮,按提示将技能包放入 AI 编程助手的 skills 目录即可。

浏览技能市场

支持平台:Qoder · QoderWork · Claude · Codex 等 AI 编程助手