Skill Guard

Skill Security Scanner - Scan for risks before download/use. Use when: installing unknown skills, evaluating third-party code, or security auditing. / Skill安...

Security Scan

The skill is internally consistent with a local static code scanner: it only requires python3, includes a simple file-scanning implementation, and does not request external credentials or perform network calls.

Purpose & Capability

Name/description match the included code: the Python module implements a pattern-based static scanner that walks a given path and searches files for suspicious tokens. Required binary (python3) is appropriate and proportional.

Instruction Scope

The SKILL.md usage shows an inspect_remote function and a RISK_PATTERNS dict that are not present in the code (the code provides scan() and check(), and RISK_PATTERNS is a list of tuples). The runtime instructions otherwise only ask you to run the scanner locally; the scanner only reads files under the provided path and does not transmit data externally.

Install Mechanism

There is no install spec that downloads external code; the repository is instruction-only with an included Python file. Nothing in the bundle writes or installs arbitrary binaries from remote URLs.

Credentials

The skill requests no environment variables or credentials. However, when you point the scanner at a directory it will read files there — so do not scan sensitive system or home directories unless intended (e.g., scanning /home/user can read files like .aws/credentials if present). This is expected for a local scanner but worth remembering.

Persistence & Privilege

Skill does not request persistent privileges, does not set always:true, and does not modify other skills or system-wide configuration. It only reads files within the supplied path.

Assessment

This appears to be a straightforward local static scanner that looks for suspicious tokens in files. Before using: (1) note that the SKILL.md and code are slightly out of sync (inspect_remote is documented but not implemented); (2) run the scanner only on the directories you intend — it will read file contents in the path you give it (don’t point it at your entire home or root unless you want those files examined); (3) expect false positives and false negatives — this is a heuristic pattern matcher, not a full taint-analysis engine; (4) because it’s local and includes no networking, it won’t exfiltrate data by itself, but always review the scanner’s source yourself if you plan to rely on its results. If you want remote inspection functionality, ask the author for an updated implementation or documentation.

🛡️ Skill Guard / 安全检查

Security scanner for AI Skills. Check code safety before install or use.

When to Use / 使用场景

EN	CN
Installing unknown skills	安装未知来源的skill
Evaluating third-party code	评估第三方代码
Security auditing	安全审计
Before running untrusted code	运行不受信任的代码前

Risk Categories / 风险类型

Category	EN	CN	Severity
Code Execution	代码执行	🔴 Critical
File Deletion	文件删除	🔴 Critical
Command Injection	命令注入	🔴 Critical
Credential Leak	凭证泄露	🟠 High
Network Request	网络请求	🟠 High
Data Theft	窃取数据	🔴 Critical
Induce Transfer	诱导转钱	🔴 Critical
Virus/Backdoor	病毒/后门	🔴 Critical

Usage / 使用

# Scan local skill / 扫描本地skill
python3 -c "from skill_guard import scan; print(scan('/path/to/skill'))"

# Inspect remote / 检查远程
python3 -c "from skill_guard import inspect_remote; print(inspect_remote('skill-name'))"

Risk Patterns / 风险模式

RISK_PATTERNS = {
    'code_execution': ['exec(', 'eval(', 'compile(', '__import__'],
    'file_deletion': ['rm -rf', 'shutil.rmtree', 'os.remove'],
    'command_injection': ['subprocess', 'os.system', 'shell=True'],
    'credential_leak': ['password', 'api_key', 'token', 'secret'],
    'data_theft': ['/etc/passwd', '~/.ssh', 'clipboard'],
    'induce_transfer': ['alipay', 'bank_card', 'payment'],
    'virus_backdoor': ['ransomware', 'miner', 'backdoor'],
}

Installation / 安装

npx clawhub install skill-guard-waai

Author / 作者

• WaaiOn