金
金融 AI 应用分级判定器
作者:注册老炮@MedXpert I SynomosAIv4
金融 AI 应用分级判定器——基于官方规则源的零依赖决策支持工具,JSON IR 输出,覆盖TH-FIN-002域。 适用场景:当需要金融 AI 应用分级判定器能力时使用本技能;触发词:合规、判定器、TH-FIN-002、fin-ai-classifier。
下载量
292
点赞
73
价格
免费
技能文档
---
name: fin-ai-classifier
slug: fin-ai-classifier
displayName: 金融 AI 应用分级判定器
version: 1.0.2
category: it-ops-security
display_name: 金融 AI 应用分级判定器
display_name_en: Financial AI Application Tiering Classifier
title: 金融 AI 应用分级判定器
author: 诺丰(SynomosAI)
license: MIT
allowed-tools: Read, Glob, Grep, Bash, Write, WebFetch
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」)
description: 金融 AI 应用分级判定器——基于官方规则源的零依赖决策支持工具,JSON IR 输出,覆盖TH-FIN-002域。 适用场景:当需要金融 AI 应用分级判定器能力时使用本技能;触发词:合规、判定器、TH-FIN-002、fin-ai-classifier。
tags: [合规, 判定器, TH-FIN-002, fin-ai-classifier, 金融AI, 分级]
description_zh: "金融 AI 应用分级判定器——基于官方规则源的零依赖决策支持工具,JSON IR 输出,覆盖TH-FIN-002域。"
description_en: "Financial AI application tiering determination tool: zero-dependency decision support with JSON IR output (domain TH-FIN-002)."
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。"
platforms: [windows, macos, linux]
read_when:
- 要对金融 AI 应用做快速合规分级(I/II/III 级),严格遵循 TH-FIN-002 官方规则
- 处理高风险金融算法决策工具的准入审查或监管报送
- 需要自动输出 JSON IR 报告接入统一监管标准接口
- 用户说「这个 AI 放贷/荐股算几级」「金融 AI 要不要备案」
ai_protocol: true
spec: usk/1.0
interface:
type: cli
entry_point: fin-ai-classifier.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: "1.0.2: Bilingual rewrite (full EN+ZH, English first), added read_when; 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
1. When a fast compliance tier assessment of a financial AI application is needed, strictly following the official rules (TH-FIN-002 domain).
2. When handling admission reviews or regulatory filings involving high-risk financial algorithmic decision tools.
3. When the system must automatically output JSON IR reports to a unified regulatory standard interface.
### Usage
Because this skill is document-type with no script dependency, no external file needs to be loaded — trigger it directly from a workflow node. When calling, state clearly the financial AI application name under classification, its core function description and its domain. Key parameters: `application_id`, `feature_description` and `domain_scope` (defaults to TH-FIN-002); the system matches the rule base automatically and returns a JSON IR result.
### Workflow
1. The user or system workflow initiates the call, passing the financial application identifier and feature profile.
2. The internal engine retrieves the official rule sources and compliance metric set within the TH-FIN-002 domain.
3. Based on the input, it auto-computes the application's risk grade and dependency level.
4. The engine produces a technical-level response conforming to the JSON IR standard.
5. The result is returned for upper-layer systems to parse, for later decisions or regulatory filing interfaces.
### Evidence & safety statement
- Evidence: TBD (no evaluation / badge link yet; update here once it is produced)
- Version: 1.0.2
- Security: no independent security scan has been performed (TBD); the conclusion is pending and no endorsement is given until it is completed. This is AI-assisted decision support, not legal advice.
## 中文
### 何时用
1. 当需对金融 AI 应用的合规性进行快速分级评估,且需严格遵循官方规则(TH-FIN-002 域)时。
2. 在处理涉及高风险金融算法决策工具的准入审查或监管报送场景时。
3. 当系统需要自动化输出 JSON IR 格式的报告以接入统一监管标准接口时。
### 用法
由于本技能为文档型且无脚本依赖,调用方式无需加载外部文件,直接通过工作流节点触发。调用时需明确待分类的金融 AI 应用名称、核心功能描述及所属领域。关键参数支持传入 `application_id`、`feature_description` 与 `domain_scope`(默认为 TH-FIN-002),系统将依据规则库进行自动匹配并返回 JSON IR 结果。
### 流程
1. 用户或系统工作流发起调用请求,传入待评级的金融应用标识及功能特征。
2. 内部引擎检索 TH-FIN-002 域内的官方规则源与合规指标集合。
3. 基于输入数据对应用的风险等级与依赖程度进行自动化匹配计算。
4. 引擎生成符合 JSON IR 标准的技术级响应数据。
5. 返回结果供上层系统解析,用于后续决策或监管报送接口对接。
### 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。本工具为 AI 辅助决策支持,非法律意见。使用说明
# fin-ai-classifier · 金融 AI 应用分级判定器 / Financial AI Application Tiering Classifier
> AI-assisted decision-support CLI. NOT legal advice — verify against official texts. Text verified: 2026-09-26.
## English
### What it does
Classifies financial AI applications into I/II/III risk grades per the **TH-FIN-002 master (FDA SaMD tiering thinking adapted to finance) + financial-AI regulatory guidance** (pending manual verification; the latest regulatory guidance prevails).
- **Insufficient input → strict presumption**: with `autonomy`/`impact` missing, the tool returns the **most obligation-heavy grade** among the possible ones, states the possible range in `warnings`, and sets `rc=1`. It never silently returns Grade I (low risk).
- **No silent invalid input**: out-of-enum values → gate error (`rc=2`).
- **Out of scope**: licensing status, filing material lists, algorithm-filing workflow.
### Install & run (zero-dependency, Python 3.8+)
```bash
python fin-ai-classifier.py --demo # 2 built-in smoke cases
# rc=1 (strict presumption, no args)
python fin-ai-classifier.py
# rc=0 (complete args → Grade I)
python fin-ai-classifier.py --autonomy info --impact low
# rc=2 (invalid input)
python fin-ai-classifier.py --autonomy full --impact extreme
```
### Parameters
| Parameter | Required | Meaning |
|---|---|---|
| `autonomy` | no* | Autonomy: full / human_in_loop / info (defaults to strict presumption) |
| `impact` | no* | Impact: high / mid / low (defaults to strict presumption) |
| `target` | no | Object: consumer / institution |
\* Semantically key tiering parameters; when missing, the tool outputs the strict-presumption conclusion with a warning, without blocking execution.
### Output
Deterministic JSON IR with `rc` (**0** = ok / **1** = completed with warnings / **2** = invalid input), `grade`, `obligations`, `warnings`, and `aigc_mark` (GB 45438-2025 metadata).
```json
{
"input": {"autonomy": null, "impact": null, "target": null},
"result": {
"grade": "Grade III (high risk)",
"obligations": ["algorithm filing", "human fallback + explainability", "model & data compliance audit", "regulatory reporting"],
"warnings": [
"Incomplete input: no autonomy / impact given; review after completing.",
"Tier uncertain (possible range Grade I–III): strict presumption took Grade III (heaviest obligations); actual grade may be lighter; review after completing parameters."
]
},
"rc": 1
}
```
### Rule-source review checklist (pending manual verification)
| Clause | Used for | Status |
|---|---|---|
| TH-FIN-002 master | I/II/III tiering framework | pending manual check |
| full + high → Grade III | Highest-grade trigger | pending manual check |
| human_in_loop + high/mid → Grade II | Mid-risk trigger | pending manual check |
| Algorithm filing (per public-opinion attributes) | Grade II obligation reference | pending manual check |
| No-licensed AI stock-recommendation / lending | Grade III warning basis | pending manual check |
### Theory hook
Financial autonomy must be tiered — tiering nails "can the AI decide for you" onto the regulatory coordinate system. Domain slot: TH-FIN-002.
## 中文
### 它做什么
依据 **TH-FIN-002 母版(FDA SaMD 分级思维平移金融)+ 金融AI监管导向**(待人工核对;以最新监管口径为准)将金融 AI 应用分为 I/II/III 风险级。
- **输入不足 → 从严推定**:缺 `autonomy`/`impact` 时返回可能区间中**义务最重的级**,在 `warnings` 中说明可能区间,置 `rc=1`,绝不静默返回 I 级(低风险)。
- **无效输入不静默**:枚举外取值 → 门禁错误(`rc=2`)。
- **不做什么**:牌照状态、备案材料清单、算法备案流程。
### 安装与运行(零依赖,Python 3.8+)
```bash
python fin-ai-classifier.py --demo # 2 个内置冒烟用例
python fin-ai-classifier.py # rc=1(从严推定,无参数)
python fin-ai-classifier.py --autonomy info --impact low # rc=0(参数齐 → I 级)
python fin-ai-classifier.py --autonomy full --impact extreme # rc=2(非法输入)
```
### 参数
| 参数 | 必填 | 含义 |
|---|---|---|
| `autonomy` | 否* | 自主度:full/human_in_loop/info(缺省按从严推定) |
| `impact` | 否* | 影响:high/mid/low(缺省按从严推定) |
| `target` | 否 | 对象:consumer/institution |
\* 语义上为分级关键参数;缺省时输出从严推定结论并告警,不阻塞执行。
### 输出
确定性 JSON IR,含 `rc`(**0**=ok / **1**=完成有告警 / **2**=非法输入)、`grade`、`obligations`、`warnings` 与 `aigc_mark`(GB 45438-2025 元数据)。
### 规则源待人工核对清单
| 条款 | 用途 | 状态 |
|---|---|---|
| TH-FIN-002 母版 | I/II/III 级分级框架 | 待人工核对 |
| full + high → III 级 | 最高级触发条件 | 待人工核对 |
| human_in_loop + high/mid → II 级 | 中风险级触发条件 | 待人工核对 |
| 算法备案(依舆论属性) | II 级义务引用 | 待人工核对 |
| 禁止 AI 荐股/放贷无资质经营 | III 级告警依据 | 待人工核对 |
### 理论钩子
金融的自治须有级——分级即把"AI 能否替你做决定"钉进监管坐标。域站位件:TH-FIN-002。
---
## License & attribution
**分层许可——代码许可不覆盖文档与理论文本。**
| 层 | 载体 | 许可 |
|---|---|---|
| 代码 | `fin-ai-classifier.py` | **MIT** |
| 文档与理论文本 | 本 `README.md`、`SKILL.md`、「理论依据」部分及一切理论表述 | **不在 MIT 覆盖范围内**:保留所有权利 |
> 代码可依 MIT 使用与再分发;文档与理论文本不因代码许可而默认可自由再发布。
```
© 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**。)
**本包补充**:`fin-ai-classifier.py` = MIT;本 `README.md` / `SKILL.md` 及其中理论文本不在任何代码许可覆盖范围内。
## 家族合集
| 域 | 工具 | 大使 |
|---|---|---|
| UAS 低空 | uas-ops-checker | 诺卫 |
| AUT 驾驶 | aut-grade-checker | 诺卫 |
| DAT 数据 | data-grade-checker | 诺源 |
| BIO 生物 | hgrac-route-checker | 诺康 |
| FIN 金融 | fin-ai-classifier | 诺丰 |
| LAW 法律 | evid-four-check | 诺律 |支持平台:Qoder · QoderWork · Claude · Codex 等 AI 编程助手