Back to skill

Security audit

Claude Code 控制器

Security checks for vulnerabilities and agentic risk

Overview

This skill is meant to control Claude Code, but it repeatedly recommends running it with permission checks bypassed for code-changing tasks.

Review before installing. Use this only in isolated project directories or disposable worktrees, avoid bypassing permission checks for routine work, do not paste secrets or sensitive logs into prompts, and do not run the recursive chmod guidance unless you have identified the exact file that needs permission changes.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (2)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:43
Finding

Claude Code Is Routinely Launched with Permission Checks Disabled

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:219
Finding

Recursive Permission Modification Grants Unnecessary Write Access Across the Project

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (10)

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
91% confidence
Finding

This skill delegates tasks to an external model provider via the claude CLI, which can transmit user prompts, code context, and repository data outside the host system. In a code-control skill, that is materially relevant because users may paste secrets, proprietary code, or logs into prompts without any warning or minimization controls.

Content

Scanner excerpt · SKILL.md (reported line 43)May include surrounding context.

bash
# 在当前目录运行
claude --permission-mode bypassPermissions --print "你的任务描述"

# 在指定项目目录运行
cd /path/to/project && claude --permission-mode bypassPermissions --print "你的任务描述"

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

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).

Content

Scanner excerpt · SKILL.md (reported line 85)May include surrounding context.

process action:list

完成后清理

git worktree remove /tmp/feature-a git worktree remove /tmp/feature-b

text

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

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).

Content

Scanner excerpt · SKILL.md (reported line 86)May include surrounding context.

完成后清理

git worktree remove /tmp/feature-a git worktree remove /tmp/feature-b

text

## 📝 常用命令模板

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
92% confidence
Finding

The code-development template sends substantial project instructions and likely source context to an external Claude CLI, creating a risk of unintentional disclosure of proprietary code or secrets embedded in task descriptions. Because the skill is specifically designed to automate development work, the chance of sharing sensitive internal context is elevated.

Content

Scanner excerpt · SKILL.md (reported line 94)May include surrounding context.

代码开发

bash
# 开发新功能
claude --permission-mode bypassPermissions --print "
开发一个用户登录功能,要求:
1. 使用 JWT 认证
2. 包含密码加密

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
93% confidence
Finding

The bug-fix template explicitly encourages pasting error logs and file paths into an external model invocation; logs often contain tokens, PII, stack traces, hostnames, or internal paths. Sending that material to a third-party CLI without screening can expose sensitive operational details.

Content

Scanner excerpt · SKILL.md (reported line 105)May include surrounding context.

md
"

# 修复 bug
claude --permission-mode bypassPermissions --print "
修复 issue #123 中的空指针异常。
错误日志:[粘贴错误信息]
相关代码:[文件路径]

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

The PR review template delegates repository diff analysis to an external provider, which may disclose proprietary source changes or security-sensitive code. While external model usage is part of the skill's purpose, the risk remains real because no consent, classification, or redaction controls are described.

Content

Scanner excerpt · SKILL.md (reported line 117)May include surrounding context.

代码审查

bash
# 审查 PR
claude --permission-mode bypassPermissions --print "
审查这个 PR 的改动:
1. 检查代码质量
2. 找出潜在 bug

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
91% confidence
Finding

The batch refactor template can expose broad swaths of a codebase to an external model service, increasing the likelihood of transmitting sensitive business logic or embedded secrets. The automation-oriented framing makes this more dangerous because users may run large-scale transformations without considering data handling implications.

Content

Scanner excerpt · SKILL.md (reported line 131)May include surrounding context.

文件操作

bash
# 批量重构
claude --permission-mode bypassPermissions --print "
将所有 .js 文件转换为 .ts:
1. 添加类型注解
2. 修复类型错误

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest description and main documentation are written entirely in Chinese and describe the skill as the dedicated interface for controlling Claude Code, without indicating that other languages are supported. This can violate a language/locale policy when a skill imposes a specific language without user opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill claims it avoids 'full permissions' but standardizes on --permission-mode bypassPermissions, which explicitly disables permission checks for the external Claude CLI. In this context, that allows autonomous code changes and potentially destructive filesystem or repo actions without interactive approval, contradicting the stated safety model.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document says sensitive operations require confirmation, but the provided command patterns invoke Claude in a mode that bypasses permission prompts and permit modification-oriented tasks such as refactoring, bug fixing, and feature development. This mismatch encourages operators to believe protections exist when the examples actually suppress them.

Content

No source excerpt is available for this finding.