Cofco Futures Skill

PassAudited by VirusTotal on May 12, 2026.

Overview

Package: cofco-futures-skill (mcp) Version: 1.0.2 Description: 中粮期货官方 AI 助手 Skill。提供公司概况查询(含分支机构)、期货交易日历查询、交割备忘录查询、开户咨询引导及客户留资服务。 The package is a Model Context Protocol (MCP) skill definition for COFCO Futures (中粮期货). It contains configuration files and documentation that define how an AI assistant should interact with official COFCO Futures services. The tools provided include company profile lookups, delivery data queries, and a workflow for account opening and futures trading via a remote server (zlqh.com). The package does not contain executable code, unexpected binary execution, or malicious data exfiltration logic. It follows standard patterns for financial service integrations, including instructions for RSA encryption of credentials and privacy warnings for PII collection.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A user may be asked to give an agent credentials for a real trading account, giving the agent delegated access to sensitive financial data and trading authority.

Why it was flagged

The subskill instructs the agent to handle trading account information and username/password login. The registry requirements declare no primary credential, and the artifacts do not clearly define credential storage, retention, or cleanup.

Skill content
有账户 | 记录用户提供的账号信息,直接进入「1. 连接与登录」流程 ... 将 "用户名&密码" 用获取的公钥加密后发送
Recommendation

Require a clearly declared, official credential flow; avoid direct password collection where possible; document storage/retention; require logout; and limit account access to the minimum necessary.

What this means

An agent could submit or cancel futures trades based on interpreted user language, which could cause financial loss if a command is misunderstood or executed without review.

Why it was flagged

The skill exposes mutable financial trading actions as raw command strings. The artifacts do not require a final user confirmation or enforce limits before order placement or cancellation.

Skill content
ws.send('order.send --instrument=rb2501 --exchange=SHFE --direction=buy --offset=open --price=4000 --volume=1'); ... ws.send('order.cancel --order_sys_id=12345');
Recommendation

Use scoped trading tools with explicit approval gates. Before every order or cancellation, show instrument, exchange, side, offset, price, volume, and risk notice, and proceed only after the user confirms.

What this means

Trading commands and account-related data may be exposed to interception or manipulation on the network.

Why it was flagged

The trading server is documented as a cleartext WebSocket endpoint. The login public key, account queries, order commands, and trade/order updates are tied to this channel; RSA login encryption alone does not authenticate the server or protect all session traffic.

Skill content
WebSocket: `ws://mcptest.zlqh.com:8765`
Recommendation

Use authenticated TLS transport such as wss://, document the production endpoint and trust boundary, and avoid sending financial account or order traffic over cleartext channels.

What this means

Users may over-trust the assistant and provide sensitive financial credentials or approve trades without independently verifying that the publisher is official.

Why it was flagged

The skill presents itself as an official financial assistant, but the supplied metadata does not provide a verifiable source or homepage. This is material because the same skill asks for trading credentials and can issue trades.

Skill content
Description: 中粮期货官方助手... Source: unknown; Homepage: none
Recommendation

Provide verifiable publisher provenance, an official homepage, and clear confirmation that the endpoints and skill are operated by the claimed financial institution.

What this means

Users will be sharing contact information for follow-up by a customer manager.

Why it was flagged

The account-opening flow collects name and phone number and submits them to a remote tool, but it does require disclosure and user confirmation, making it purpose-aligned rather than inherently suspicious.

Skill content
询问用户姓名 ... 询问用户联系电话 ... 展示给用户,让用户确认 ... 调用 `collect_customer_info` 工具提交数据
Recommendation

Only provide contact information if you want follow-up, and confirm the displayed name and phone number before submission.