数学表达式计算器。支持四则运算、科学函数(sin/cos/tan/log/sqrt)、常量(pi/e)、单位换算。安全求值不使用eval，纯Python标准库，无需API Key。

This skill is consistent with its stated purpose, but exercise caution before using it on untrusted or security-sensitive systems. The script uses eval() after transforming the input; although builtins are cleared and a restricted namespace is provided, the implementation has two practical concerns: - Correctness risk: it replaces the character 'e' with Euler's constant (string replace), which will break scientific notation like "2e3" and may silently produce incorrect results. - Safety/abuse risk: eval() is used (even with restricted globals). The allowed-character check still permits '.' and '_' which enable attribute access in expressions; while other syntactic elements (like brackets) are disallowed, subtle attack vectors or future code changes could expand risk. Recommendations before installing or invoking broadly: - If you control the code: replace eval-based evaluation with a proper parser (for example a dedicated expression parser or an AST-based whitelisting evaluator). Avoid global replacement of single letters for constants; parse tokens instead. - If you don't control the code: test edge cases (e.g., "2e3", complex inputs), run the script in an isolated environment, and avoid exposing it to untrusted user input or enabling the skill as always-on. - If you need a high-assurance calculator, prefer implementations that use a grammar-based parser or safe-eval libraries and have been audited. Confidence is medium because the script is mostly coherent, but the use of eval and the 'e' replacement are concrete, non-trivial issues that justify a cautious classification.

Type: OpenClaw Skill Name: cn-math-calculator Version: 1.1.0 The skill is classified as suspicious due to a critical Remote Code Execution (RCE) vulnerability and misleading documentation. Although SKILL.md explicitly claims the tool does not use 'eval', the implementation in scripts/math_calculator.py uses eval() to process mathematical expressions. The input validation regex allows characters like '.' and '_', which are sufficient to perform Python attribute access (e.g., __class__) and bypass the intended sandbox, potentially allowing arbitrary code execution. While there is no clear evidence of intentional malice or exfiltration, the contradiction between the documentation and the insecure code is a significant red flag.