Skill Auditor Pro

审查 ClawHub skills 的安全性，检测恶意代码、可疑模式和社工攻击。在安装任何第三方 skill 前使用此工具进行安全检查。

This skill appears to implement a reasonable audit workflow, but review these points before installing or running it: 1) Ensure the host has 'clawhub' (and optionally the 'gemini' CLI) if you plan to use fetch or LLM analysis — the skill does not declare these requirements. 2) Be cautious with the L3 step: the script copies 'suspicious' code to /tmp and instructs an agent/LLM to analyze it — that can leak secrets or sensitive code to whatever model/service you invoke. Prefer running the audit locally and reviewing findings yourself rather than sending raw suspicious files to remote LLMs or public APIs. 3) Confirm the external reporting address ([email protected]) and any IoC lists are from a trusted source before using them. 4) If you need to proceed: run the script in an isolated environment, inspect its output and the /tmp artifact contents manually, and only use LLM analysis on redacted content that removes any credentials or secrets. If you want a safer verdict, provide: where this skill was published (homepage/source repo), whether 'clawhub' and 'gemini' are available in your environment, and whether you intend to allow the agent to send suspicious code to external LLM services.

Type: OpenClaw Skill Name: skill-auditor-pro Version: 2.1.1 This skill is a security scanner designed to audit other OpenClaw skills. Its code and documentation are entirely consistent with this stated purpose. It implements pattern matching, deobfuscation, and LLM-based intent analysis to detect malicious patterns (e.g., `base64|bash`, `curl|bash`, `eval`, access to sensitive directories like `~/.ssh` or `~/.openclaw`, credential keywords, known C2 IPs/domains). The script creates a temporary file containing identified suspicious code and instructs the OpenClaw agent to `cat` this file for LLM analysis, which is a functional interaction aligned with its 'L3: LLM Intent Analysis' feature, not a malicious prompt injection attempt.