Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed permission-checking guide, and the flagged dangerous commands appear as examples of what to block rather than instructions to execute.
Before installing, be aware this skill is guidance-only and broad: it may influence how an agent asks about or auto-classifies shell commands. Review any local permission rules you add, avoid using bypass mode outside tests, and keep enterprise or platform permission controls authoritative.
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
eval, exec, xargs, sudo
// 管道执行 curl | bash, wget | sh
### 危险模式检测
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
eval, exec, xargs, sudo
// 管道执行 curl | bash, wget | sh
### 危险模式检测
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
eval, exec, xargs, sudo
// 管道执行 curl | bash, wget | sh
### 危险模式检测
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
eval, exec, xargs, sudo
// 管道执行 curl | bash, wget | sh
### 危险模式检测
Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.
eval, exec, xargs, sudo
// 管道执行 curl | bash, wget | sh
### 危险模式检测
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
{
"dangerousCommands": [
"rm -rf /",
"dd if=* of=/dev/*",
":(){ :|:& };:" // Fork bomb
]
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
{
"dangerousCommands": [
"rm -rf /",
"dd if=* of=/dev/*",
":(){ :|:& };:" // Fork bomb
]
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
{
"dangerousCommands": [
"rm -rf /",
"dd if=* of=/dev/*",
":(){ :|:& };:" // Fork bomb
]
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
{
"dangerousCommands": [
"rm -rf /",
"dd if=* of=/dev/*",
":(){ :|:& };:" // Fork bomb
]
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
{
"dangerousCommands": [
"rm -rf /",
"dd if=* of=/dev/*",
":(){ :|:& };:" // Fork bomb
]
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
{
"dangerousCommands": [
"rm -rf /",
"dd if=* of=/dev/*",
":(){ :|:& };:" // Fork bomb
]
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
"defaultMode": "auto",
"dangerousCommands": [
"rm -rf /",
"rm -rf /home/*",
":(){ :|:& };:"
],
"autoAllow": [
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
"defaultMode": "auto",
"dangerousCommands": [
"rm -rf /",
"rm -rf /home/*",
":(){ :|:& };:"
],
"autoAllow": [
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
],
"autoDeny": [
"curl | bash",
"wget -O- | sh"
],
"rules": [
{
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
🚫 Auto-denied: rm -rf /*
Reason: 危险命令，已拦截
The manifest description and the entire user-facing skill documentation are written in Chinese, including prompts and response formats, with no indication that users may choose another language. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale constraint is documented and justified.
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
// Shell
bash, sh, zsh, fish, ssh
eval, exec, xargs, sudo
// 管道执行
curl | bash, wget | sh
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
ls, ll, cat, head, tail, grep, find, pwd
// 目录操作
cd, mkdir -p, tree, stat, file
// 开发工具
npm --version, node --version, git --version
No suspicious patterns detected.