Security audit
Security checks for vulnerabilities and agentic risk
This is a local security-audit skill whose file scanning and limited fix actions match its stated purpose, but users should run fix mode deliberately because it changes local files.
Install only if you want a Chinese-language local audit helper with read and exec access. Review results carefully because reports may include snippets of detected secrets, and run fix mode only in the intended project directory since it can change .env permissions and create .gitignore.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
"""检查文件权限"""
target_path = Path(target)
# ...
sensitive_files = [".env", ".env.local", ".env.production", "id_rsa", "config.json"]
for filename in sensitive_files:
filepath = target_path / filename
stat = filepath.stat()
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
"""检查文件权限"""
target_path = Path(target)
# ...
sensitive_files = [".env", ".env.local", ".env.production", "id_rsa", "config.json"]
for filename in sensitive_files:
filepath = target_path / filename
stat = filepath.stat()
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
"""检查文件权限"""
target_path = Path(target)
# ...
sensitive_files = [".env", ".env.local", ".env.production", "id_rsa", "config.json"]
for filename in sensitive_files:
filepath = target_path / filename
stat = filepath.stat()
Potential security issue detected. Manual review is recommended.
"file": str(dockerfile),
"fix": "添加非root用户: RUN useradd -m appuser && USER appuser"
})
if "--privileged" in content:
"level": "HIGH",
"category": "docker",
"message": "Docker容器使用特权模式",
Potential security issue detected. Manual review is recommended.
"file": str(dockerfile),
"fix": "添加非root用户: RUN useradd -m appuser && USER appuser"
})
if "--privileged" in content:
"level": "HIGH",
"category": "docker",
"message": "Docker容器使用特权模式",
The auto-fix routine unconditionally modifies the local .env file permissions when present, without confirmation, backup, or target scoping. While intended as hardening, any state-changing operation on credential-bearing files in an agent-executed skill is risky because it can affect developer workflows, break shared environments, or alter files outside the user's intended context.
fixes = []
# ...
# 修复.env权限
env_file = Path(".env")
if env_file.exists():
os.chmod(env_file, 0o600)
fixes.append("已设置.env文件权限为600")
SKILL.md 在 L039 明确将该技能描述为“纯本地执行,无需外部依赖”。但代码在 L282-L284 通过 subprocess.run(["ss", "-tlnp"]) 依赖外部系统工具 ss，且依赖说明 L475 也承认这一点，因此文档表述与实现相矛盾。
The skill advertises an automatic fix mode that changes file permissions and creates files, but it does not require an explicit warning, dry-run, or user confirmation before making local changes. In an agent context, this can lead to unintended modification of the user's workspace and security-relevant files, especially if the skill is triggered implicitly or run against the wrong directory.
The trigger conditions are broad enough that the skill may activate on generic security-related requests and then invoke exec-backed audit or fix behavior unexpectedly. In an agent environment, overbroad triggering increases the chance of unintended local scanning, command execution, or file modification without the user explicitly asking for this specific skill.
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
"category": "permissions",
"message": f"敏感文件权限过宽: {filename} ({perms})",
"file": str(filepath),
"fix": f"执行: chmod 600 {filename}"
})
# ...
def _check_docker(self, target):
L005-L006以及全文说明均以中文固定呈现，但文档未说明这是面向特定中文用户群的地域化技能，也未提供语言选择或用户 opt-in。对于支持多平台和广泛 Agent 的技能，固定单一语言可能构成语言/地区策略上的自然语言约束。
L453 的说明把自动修复范围描述为包括“加固配置文件”。实际 _auto_fix 实现在 L365-L374 只执行 chmod .env 和创建 .gitignore，没有任何配置文件加固逻辑，因此文档主动夸大了代码行为。
No suspicious patterns detected.