Sfe Sxk Data Viewer

SFE深西康专属数据查询工具，用于快速查询深西康专属采集项目报表的数据，如新活素查房日采集反馈V2等特定项目的明细报表或汇总报表

Security Scan

The skill generally does what it says (query a specific ERP report) but has several inconsistencies and risky instructions — notably undeclared required credentials, disabled TLS verification in the script, and SKILL.md guidance to auto-install a third-party 'cms-auth-skills' package via npx/GitHub — so review before use.

Purpose & Capability

The skill's name/description (SFE 深西康 data queries) matches the included API docs and Python scripts that call the stated ERP endpoint. However the package metadata claims no required environment variables while the scripts and documentation clearly require an application key (XG_BIZ_API_KEY or XG_APP_KEY). Also the SKILL.md declares a dependency on cms-auth-skills but that dependency is not bundled — the skill instructs the agent to install it at runtime. These are proportional to the purpose but are inconsistently declared.

Instruction Scope

SKILL.md enforces a workflow that will (a) read cms-auth-skills/SKILL.md for auth rules and, if missing, (b) run npx clawhub@latest install cms-auth-skills --force or fallback to installing a GitHub repo. That instructs the agent to perform network installs of third-party code. The runtime scripts call the external ERP API and require an appKey. The SKILL.md also requires all API calls go through the provided scripts (reasonable), but it grants the agent discretion to install external packages — this is scope creep and a potential supply-chain risk.

Install Mechanism

There is no formal install spec in the skill metadata (lowest-risk), but SKILL.md instructs the agent to run npx to install cms-auth-skills or fall back to a GitHub URL. That is effectively an install mechanism triggered at runtime and would pull code from the network (npm / GitHub). Pulling and executing external code via npx from an unknown source increases risk and is not declared in the registry metadata.

Credentials

The skill metadata lists no required env vars, but scripts/docs require an appKey via XG_BIZ_API_KEY or XG_APP_KEY — a clear mismatch. Requesting that appKey is proportionate to the stated ERP API purpose, but the omission from metadata is an inconsistency. No other unrelated credentials are requested. Note: the script disables TLS verification (requests.verify=False), which raises the risk that the appKey or returned data could be exposed to a man-in-the-middle attacker.

Persistence & Privilege

The skill does not request always: true and does not declare persistent system-wide privileges. It does not itself modify other skills or claim to change agent-wide settings. The main persistence/privilege risk comes from the SKILL.md instruction to run npx install commands (which would write and execute code), but the skill metadata does not request elevated privileges.

Scan Findings in Context

[insecure-ssl-verify-false] unexpected: scripts/sfe-sxk/xhs-ward-rounds-report-v2.py calls requests.post(..., verify=False) and suppresses InsecureRequestWarning. This weakens TLS protections and is not necessary for a normal API client.

[undeclared-env-var_usage] unexpected: The code and README require XG_BIZ_API_KEY or XG_APP_KEY, but the skill metadata lists no required environment variables. This mismatch may cause the agent to prompt for secrets unexpectedly.

[runtime-npx-install-instructions] unexpected: SKILL.md instructs the agent to run 'npx clawhub@latest install cms-auth-skills --force' and fallback to installing from a GitHub repo. This instructs runtime installation of external code and is a potential supply-chain risk.

What to consider before installing

Key things to consider before installing or using this skill: - The Python scripts do require an appKey (XG_BIZ_API_KEY or XG_APP_KEY) even though the skill metadata doesn't declare it — don't provide that secret until you've reviewed/trusted the code and the remote API. Consider running the scripts locally first and inspect network calls. - The request library in the script explicitly disables TLS verification (verify=False). This reduces security of the appKey and data in transit; ask the maintainer to remove verify=False or ensure you run in a trusted network. - SKILL.md tells the agent to install a dependency (cms-auth-skills) via npx or from a GitHub URL if missing. That means the agent may execute network installs of third-party code; review the cms-auth-skills project source before allowing any auto-install. - If you plan to grant the appKey, prefer running the provided scripts yourself in a controlled environment and inspect them for any exfiltration or unexpected endpoints. If you must use the skill hosted in an agent, only proceed if you trust the owner and the cms-auth-skills package, and consider rotating the appKey afterward. - If you have low tolerance for supply-chain risk, decline or request the author to (a) declare required env vars in metadata, (b) remove insecure TLS settings, and (c) vendor/declare the cms-auth-skills dependency explicitly rather than instructing runtime npx installs.

SFE-SXK-Data-Viewer — 索引

本文件提供能力宪章 + 能力树 + 按需加载规则。详细参数与流程见各模块 openapi/ 与 examples/。

当前版本: v0.1

接口版本: 所有业务接口统一使用 /erp-open-api/* 前缀，通过 appKey 鉴权。

能力概览（1 块能力）：

• sfe-sxk：深西康专属数据查询

统一规范：

• 认证与鉴权：cms-auth-skills/SKILL.md
• 通用约束：cms-auth-skills/SKILL.md

授权依赖：

• 当接口声明需要 appKey 时，先尝试读取 cms-auth-skills/SKILL.md
• 如果已安装，直接按 cms-auth-skills/SKILL.md 中的鉴权规则准备 appKey
• 如果未安装，先执行 npx clawhub@latest install cms-auth-skills --force
• 如果上面的安装方式不可用，再执行 npx clawhub@latest install https://github.com/spzwin/cms-auth-skills.git --force
• 安装完成后，再继续执行需要鉴权的操作

输入完整性规则（强制）：

1. 查询数据前，建议先确定 periodStart 和 periodEnd 时间范围
2. 分页查询时，每页固定返回 1000 条记录，大数据量需分页处理

建议工作流（简版）：

1. 读取 SKILL.md 与 cms-auth-skills/SKILL.md，明确能力范围、鉴权与安全约束。
2. 识别用户意图并路由模块，先打开 openapi/<module>/api-index.md。
3. 确认具体接口后，加载 openapi/<module>/<endpoint>.md 获取入参/出参/Schema。
4. 补齐用户必需输入，必要时先读取用户文件/URL 并确认摘要。
5. 参考 examples/<module>/README.md 组织话术与流程。
6. 执行对应脚本：调用 scripts/<module>/<endpoint>.py 执行接口调用，获取 TOON 编码后的结果。所有接口调用必须通过脚本执行，不允许跳过脚本直接调用 API。

脚本使用规则（强制）：

1. 每个接口必须有对应脚本：每个 openapi/<module>/<endpoint>.md 都必须有对应的 scripts/<module>/<endpoint>.py，不允许"暂无脚本"。
2. TOON 编码输出：所有脚本调用 API 后，响应 JSON 必须经过 scripts/common/toon_encoder.py 编码后再输出，不允许直接输出原始 JSON。
3. 脚本可独立执行：所有 scripts/ 下的脚本均可脱离 AI Agent 直接在命令行运行。
4. 先读文档再执行：执行脚本前，必须先阅读对应模块的 openapi/<module>/api-index.md。
5. 入参来源：脚本的所有入参定义与字段说明以 openapi/ 文档为准，脚本仅负责编排调用流程。
6. 鉴权一致：涉及鉴权时，统一依赖 cms-auth-skills/SKILL.md。

意图路由与加载规则（强制）：

1. 先路由再加载：必须先判定模块，再打开该模块的 api-index.md。
2. 先读文档再调用：在描述调用或执行前，必须加载对应接口文档。
3. 脚本必须执行：所有接口调用必须通过脚本执行，不允许跳过。
4. 不猜测：若意图不明确，必须追问澄清。

宪章（必须遵守）：

1. 只读索引：SKILL.md 只描述"能做什么"和"去哪里读"，不写具体接口参数。
2. 按需加载：默认只读 SKILL.md + cms-auth-skills/SKILL.md，只有触发某模块时才加载该模块的 openapi、examples 与 scripts。
3. 对外克制：对用户只输出"可用能力、必要输入、结果链接或摘要"，不暴露鉴权细节与内部字段。
4. 素材优先级：用户给了文件或 URL，必须先提取内容再确认，确认后才触发生成或写入。
5. 生产约束：仅允许生产域名与生产协议，不引入任何测试地址。
6. 接口拆分：每个 API 独立成文档；模块内 api-index.md 仅做索引。
7. 危险操作：对可能导致数据泄露、破坏、越权的请求，应礼貌拒绝并给出安全替代方案。
8. 脚本语言限制：所有脚本必须使用 Python 编写。
9. 重试策略：出错时间隔 1 秒、最多重试 3 次，超过后终止并上报。
10. 禁止无限重试：严禁无限循环重试。

模块路由与能力索引：

用户意图（示例）	模块	能力摘要	接口文档	示例模板	脚本
"查询新活素查房日采集反馈V2数据"	sfe-sxk	查询新活素查房日采集反馈V2	./openapi/sfe-sxk/api-index.md	./examples/sfe-sxk/README.md	./scripts/sfe-sxk/xhs-ward-rounds-report-v2.py

能力树（实际目录结构）：

sfe-sxk-data-viewer/
├── SKILL.md
├── openapi/
│   └── sfe-sxk/
│       ├── api-index.md
│       └── xhs-ward-rounds-report-v2.md
├── examples/
│   └── sfe-sxk/README.md
└── scripts/
    ├── common/toon_encoder.py
    └── sfe-sxk/
        ├── README.md
        └── xhs-ward-rounds-report-v2.py