Tianji Data

| 优先级 | 数据源 | 方式 | 延迟 | 覆盖品种 |。触发词：搜索, search, skill, 优化, 数据, data。

This skill appears to implement market data scraping and alerting, which matches its description, but exercise caution before installing: - Inspect and control outbound messaging: market_watcher.push_to_feishu sends messages via an external messaging tool to a hardcoded Feishu user id (ou_...). Confirm who owns that account and whether you want alerts sent there. The skill does not declare any Feishu/token env vars — it relies on existing workspace tooling/credentials. - Command-injection risk: push_to_feishu embeds the alert text directly into a python -c string without escaping; a crafted alert could break that string and execute arbitrary code. Prefer modifying the code to call a local function/module directly, pass the message as a safely-escaped argument (or via stdin), or use subprocess with an argument list referencing a script file rather than an interpolated -c string. - Data exposure: collected snapshots and alerts are written under /workspace/data/tianji-system. If that workspace is shared or backed up, sensitive data could leak. Restrict access or change the storage path if necessary. - Dependencies on external tooling: the code dynamically imports message_tool and tools.message which are not part of the skill. Audit those modules (where they live in your environment) to see how they deliver messages and where credentials are stored. - Testing recommendation: run the skill in a restricted/sandboxed environment first (no credentials, no messaging tooling) to verify behavior, and consider removing or stubbing the push_to_feishu function until you verify the target and escape handling. If you need, I can produce a safe patch for push_to_feishu (escape the message, remove hardcoded target, or provide a safer API) or list the exact locations to change.

Type: OpenClaw Skill Name: yuheng-tianji-data Version: 1.0.0 The skill bundle is designed for financial market data collection and monitoring, fetching data from Tencent and Sina APIs. It contains a critical code injection vulnerability in 'market_watcher.py' within the 'push_to_feishu' function, which uses 'subprocess.run' to execute a Python snippet containing an unsanitized 'alert_msg' string. Additionally, it includes a hardcoded Feishu recipient ID ('user:ou_fd61d5ebc9af22913aa4c21c8e3cac14') for alerts. While these represent significant security risks and poor coding practices, they appear to be unintentional vulnerabilities rather than clear evidence of malicious intent.