Medical Device Recall-level Decision Checker

技能文档

---
name: recall-decision-checker
title: "Medical Device Recall-level Decision Checker"
slug: recall-decision-checker
display_name: 医疗器械召回级别判定器
display_name_en: Medical Device Recall-level Decision Checker
version: 1.0.1
category: medtech-reg
author: SynomosAI
license: MIT
allowed-tools: Read, Glob, Grep, Bash, Write, WebFetch
description: 医疗器械召回级别判定器——基于官方规则源的零依赖决策支持工具，JSON IR 输出，覆盖 TH-MED-003 域（医疗器械全链路）。
description_zh: "医疗器械召回级别判定器——基于官方规则源的零依赖决策支持工具，JSON IR 输出，覆盖 TH-MED-003 域（医疗器械全链路）。"
description_en: "Medical device recall-level determination tool: zero-dependency decision support with JSON IR output (domain TH-MED-003, full device chain)."
tags: [医疗器械, 合规, 判定器, TH-MED-003, recall-decision-checker, medtech-reg, 召回]
classification:
  skillhub: ["medtech-reg", "medical-device", "recall"]
  clawhub: ["productivity", "development"]
  iso_25010: ["Functional suitability"]
read_when:
  - 需依据 NMPA 规则对召回风险分级且判定须基于官方规则源
  - 全链路质量管理需结构化 JSON IR 结果对接 ERP/QMS
  - 复杂产品缺陷报告需不依赖本地脚本的合规自动判定
risk_tier: medium
risk_rationale: "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。"
platforms: [windows, macos, linux]
platform_compatibility: [windows, macos, linux]
ai_protocol: true
spec: usk/1.0
interface:
  type: prompt
  entry_point: SKILL.md
  runtime: none
  call_pattern: prompt
output_schema:
  stdout: structured JSON IR
  exit_code: rc=0 ok / rc=1 warning / rc=2 bad input
changelog: "1.0.1: Bilingual rewrite (EN-first) + read_when; risk_rationale ref neutralized; 1.0.0 initial"
---

> 凡自治之物 · LGD 治理学 · 创造术家族 ｜ 评测协议：skill-quality-gate

## English

# Medical Device Recall-level Decision Checker

## When to use
1. When recall risk must be graded under NMPA rules with logic based on official sources, not an external model.
2. In full-chain quality management, when structured JSON IR results must feed directly into ERP/QMS interfaces.
3. For complex defect reports requiring automated compliance determination without local scripts.

## How to use
Input parameters as JSON: product model, defect description, affected-population scope. No scripts; reference TH-MED-003 rules as the reasoning basis and request JSON IR output with recall level, regulatory clauses, and confidence score.

## Workflow
1. Receive submitted data and validate completeness against TH-MED-003.
2. Extract key product attributes and retrieve the recall-level grading matrix.
3. Logically compare against risk indicators and affected scope with weighting.
4. Generate standard JSON IR with the recommended recall level.
5. Attach regulatory references for manual review and traceability.

## Evidence & safety
- Evidence: TBD; Version: 1.0.1; Security: no independent scan (TBD).

## 中文

# 医疗器械召回级别判定器

## 何时用
1. 需依据 NMPA 规则对召回风险分级，且判定逻辑须基于官方规则源而非外部模型。
2. 全链路质量管理场景，需结构化 JSON IR 结果直接对接 ERP/QMS 接口。
3. 复杂产品缺陷报告，需不依赖本地脚本的合规自动判定查询。

## 用法
以 JSON 传入产品型号、缺陷描述、受影响群体范围；无脚本，对话中引用 TH-MED-003 规则作为推理依据，要求返回含召回等级、法规条款、置信度评分的 JSON IR。

## 流程
接收数据并校验完整性 → 提取产品属性并检索召回级别判定矩阵 → 按风险指标与受影响范围做权重比对 → 生成含推荐召回级别的 JSON IR → 附加法规依据索引供人工复核与追溯。

## 证据与安全声明
- 证据：TBD；版本：1.0.1；安全：未做独立安全扫描（TBD）。

使用说明

# recall-decision-checker · 医疗器械召回级别判定器

[![LGD](https://img.shields.io/badge/LGD-Governance-blue)](https://github.com/zhaoxinghua09-cell/lgd-theory)

> AI-assisted decision-support CLI. NOT legal advice, **not a compliance proof** — verify against official texts. Text verified: 2026-09-26.

## What it does
Determines the **recall level and notification deadline** per official rules: **《医疗器械召回管理办法》(总局令 第29号)** — based on `harm_level` (severe/reversible/none, required) and `use_phase` (on_market/in_use/sold, optional).

- 一级（severe）→ 1 日内通知；二级（reversible）→ 3 日内；三级（none）→ 7 日内；均须向省局报告。
- **Insufficient input → strict presumption**: `use_phase=in_use/sold` → warning that the notification scope must also cover distributors and users, and in-use units need follow-up per the recall plan (`rc=1`). It never silently skips in-market obligations.
- Invalid enum values are rejected (`rc=2`) — garbage `harm_level` is NOT silently treated as Level III.

## Install & run (zero-dependency, Python 3.8+)
```bash
python recall-decision-checker.py --demo
python recall-decision-checker.py --harm_level severe                 # Level 1, 1-day, rc=1
python recall-decision-checker.py --harm_level reversible             # Level 2, 3-day, rc=0
python recall-decision-checker.py --harm_level severe --use_phase sold  # in-market notice warning, rc=1
python recall-decision-checker.py --harm_level garbage                # invalid enum → rc=2
```

## Output
Deterministic JSON IR with `rc` (0=ok / 1=with warnings / 2=insufficient or invalid input), `recall_level`, `notify_within_days`, `obligations`, `warnings`, `evidence`, and `aigc_mark` (GB 45438-2025 metadata, `is_generated: true`). Decision support only — final determination rests with the regulator.

## Compliance & IP
- LGD-Powered badge · theory TH-MED-003 · ambassador 诺康@MED 域首席发声人
- 中文文档见 `SKILL.md`。本工具为决策支持，须人工复核，最终以监管机构认定为准。

---

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

| 层 | 载体 | 许可 |
|---|---|---|
| 代码 | `recall-decision-checker.py` | **未在 LICENSE 中识别** |
| 文档与理论文本 | 本 `SKILL.md`、`README.md`、其中「理论依据」段与一切理论表述 | **不在代码许可覆盖范围内**：保留所有权利（All rights reserved） |

> 即：**代码可依其自身许可使用与再分发；文档与理论文本不因代码许可而默认可自由再发布。**

```
© 2026 SynomosAI. 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：**A 段公共段逐字复制、未删改；B 段资产参数段按本包真值填写**；发布 / 重提前须照最新版本复核。）

## 家族合集
| 环节 | 工具 | 大使 |
|---|---|---|
| 非临床研究 | non-clinical-checker | 诺康 |
| 上市后不良事件 | pms-vigilance-intake | 诺康 |
| 召回 | recall-decision-checker | 诺康 |
| 标签 | label-compliance-checker | 诺康 |
| 进出口 | import-export-router | 诺衡 |

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。