中文 EN

扫描报告

扫描新文件

✓ 可信 — 风险评分 5/100
上次扫描：20 小时前 重新扫描
5 /100
Agent Notify
Cross-platform notification sound and taskbar flash for AI coding agents
Agent Notify is a legitimate, well-documented notification utility for AI coding agents. All behavior — shell execution, file writes, agent config modification — is explicitly declared in SKILL.md. No malicious patterns, credential access, data exfiltration, or obfuscation observed.
技能名称Agent Notify
分析耗时43.8s
引擎pi
✓
可以安装
No action required. This skill is safe to use.

安全发现 3 项

严重性 安全发现 位置
低危
Overly broad trigger keywords 文档欺骗
The skill defines 26 Chinese and English trigger keywords (e.g., '提示音', 'notify', 'beep', 'bell'). While expansive, this is a common pattern for skill engines and is clearly documented with no security impact.
triggers: ["提示音", "通知提醒", ..., "notification setup"]
→ Consider narrowing trigger keywords if false activations become problematic. This is a UX concern, not a security issue.
 skill.md:1
低危
Missing Windows script file 供应链
SKILL.md references scripts/notify-windows.ps1 and the README project structure lists it, but the file is not present in the package. If a user selects Windows during quick setup, the installation would fail or attempt an incomplete configuration.
cp "<skill_dir>/scripts/notify-windows.ps1" "$AGENT_HOME/notify.ps1"
→ Add the missing notify-windows.ps1 script or remove Windows from supported platforms in the documentation.
 skill.md:88
提示
No dependency pinning 供应链
The skill relies on external tools (jq, python3, paplay, aplay, pw-play, notify-send, afplay) without pinning versions. These are standard OS-level tools, not package dependencies, so this is low risk.
if command -v jq &>/dev/null; then
→ No action required. These are system binaries, not pip/npm packages.
 scripts/notify-linux.sh, scripts/notify-macos.sh:1
资源类型声明权限推断权限状态证据
文件系统 WRITE WRITE ✓ 一致 skill.md:3 — 'Skill definition' with full install/uninstall instructions for fil…
命令执行 WRITE WRITE ✓ 一致 skill.md:3 — PowerShell/bash execution explicitly documented for OS detection, s…
网络访问 NONE NONE — No network operations in any script
环境变量 NONE NONE — No environment variable access in any script
剪贴板 NONE NONE — No clipboard access
浏览器 NONE NONE — No browser access
数据库 NONE NONE — No database access
技能调用 NONE NONE — No cross-skill invocation
3 项发现
🔗
中危 外部 URL 外部 URL
https://img.shields.io/badge/License-GPLv3-blue.svg
README.md:7
🔗
中危 外部 URL 外部 URL
https://www.gnu.org/licenses/gpl-3.0
README.md:7
🔗
中危 外部 URL 外部 URL
https://img.shields.io/badge/Platform-Windows%20%7C%20macOS%20%7C%20Linux-brightgreen.svg
README.md:8

目录结构

5 文件 · 21.5 KB · 759 行
Markdown 2f · 588L Shell 2f · 148L JSON 1f · 23L
├─ ▾ 📁 config
│ └─ 📋 default.json JSON 23L · 427 B
├─ ▾ 📁 scripts
│ ├─ 🔧 notify-linux.sh Shell 76L · 2.1 KB
│ └─ 🔧 notify-macos.sh Shell 72L · 2.1 KB
├─ 📝 README.md Markdown 165L · 4.2 KB
└─ 📝 skill.md Markdown 423L · 12.7 KB

安全亮点

✓ All file operations are explicitly declared in SKILL.md with exact commands and file paths
✓ Agent config modification uses safe merge strategy — never overwrites existing settings
✓ No credential harvesting, API key access, or sensitive path traversal
✓ No base64 encoding, eval, or obfuscated execution patterns
✓ No network requests or data exfiltration endpoints
✓ Scripts use graceful fallbacks for missing tools (e.g., jq→python3, paplay→aplay→pw-play)
✓ Uninstall function properly cleans all created files and settings entries
✓ OS detection via uname is non-destructive and read-only
✓ Config parsing uses standard JSON libraries with no shell injection vectors