macos-security-scan 安全扫描报告 — 可信 | ClawSafe

5 /100

macos-security-scan

Read-only macOS security scan for malware, keyloggers, suspicious processes, network connections, startup items, and privacy permissions

A legitimate, read-only macOS security scanner with clean code, no obfuscation, no credential access, no network exfiltration, and all capabilities properly used for the stated purpose.

技能名称macos-security-scan

分析耗时68.8s

引擎pi

✓

可以安装

This skill is safe to use. No security concerns identified.

安全发现 2 项

严重性	安全发现	位置
低危	Minor imprecision: 'read-only' claim vs. file write 文档欺骗 SKILL.md states 'It is safe to run — it only reads system state and never modifies anything' but the script writes a report file to disk. The file write is declared via the --out argument in SKILL.md documentation, and the write does not modify system state (only creates a new report). This is a minor doc imprecision rather than deceptive behavior. `This skill runs a comprehensive, read-only security scan of a macOS machine and produces a detailed report. It is safe to run — it only reads system state and never modifies anything.` → Update SKILL.md to clarify: 'read-only system scan that writes its output to a report file.'	`SKILL.md:12`
低危	Sudo flag ('-n' non-interactive) not explicitly called out in SKILL.md permissions 权限提升 The script uses 'sudo -n' (non-interactive sudo) when --sudo is passed. While SKILL.md mentions 'some checks need sudo for full results' in the permissions field, the non-interactive nature of '-n' isn't called out. `cmd = ["sudo", "-n"] + cmd` → Consider adding a note in SKILL.md that --sudo uses non-interactive sudo which requires prior sudo privileges without password prompt.	`scripts/scan.py:72`

资源类型	声明权限	推断权限	状态	证据
命令执行	`WRITE`	`WRITE`	✓ 一致	scripts/scan.py:74 — subprocess.run() used only for read-only system commands (p…
文件系统	`READ`	`WRITE`	✓ 一致	scripts/scan.py:465-467 — writes report to --out path (default ~/Desktop/securit…
网络访问	`NONE`	`NONE`	—	No network requests made. socket.gethostname() is local-only system info.
环境变量	`NONE`	`NONE`	—	No os.environ access or credential reading. No sensitive path access.
剪贴板	`NONE`	`NONE`	—	No clipboard access.
浏览器	`NONE`	`READ`	✓ 一致	scripts/scan.py:357-378 — reads Chrome, Firefox, Safari extension directories. D…
数据库	`NONE`	`READ`	✓ 一致	scripts/scan.py:302-315 — sqlite3 reads TCC.db (privacy database). Declared in S…

目录结构

2 文件 · 23.4 KB · 610 行

Python 1f · 479L Markdown 1f · 131L

├─ ▾ 📁 scripts

│ └─ 🐍 scan.py Python 479L · 17.9 KB

└─ 📝 SKILL.md Markdown 131L · 5.4 KB

依赖分析 1 项

包名	版本	来源	已知漏洞	备注
`None (stdlib only)`	`N/A`	built-in	否	No external dependencies. Uses only Python 3 standard library (argparse, datetime, json, os, platform, plistlib, re, socket, subprocess, sys, pathlib).

安全亮点

✓ No obfuscation: no base64, eval(), atob(), or other code obfuscation techniques

✓ No credential theft: no access to ~/.ssh, ~/.aws, .env, or iteration over os.environ

✓ No data exfiltration: no outbound network requests, no POSTs, no external IP connections

✓ No reverse shell or C2: no socket-based remote execution, no suspicious network listeners

✓ No malicious supply chain: no external dependencies (0 pinned packages), only Python stdlib used

✓ No hidden functionality: all check categories are documented in SKILL.md

✓ Command injection safe: all subprocess commands are hardcoded as static lists, no string interpolation

✓ No persistence mechanisms: no cron, no startup hooks, no backdoor installation

✓ Timeout protection: all subprocess calls have a 15-second timeout

✓ OS platform guard: script explicitly exits with error on non-Darwin platforms

✓ Conservative malware patterns: known-bad list is short and well-documented

✓ SKILL.md accurately maps all check categories to the skill's behavior

✓ No suspicious file/directory access patterns