ClawHub
Back to skill

Security audit

neat

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed knowledge-cleanup skill that reads and updates project docs and agent memory, with meaningful guardrails for destructive changes.

Install this only if you want an agent to reconcile and edit project documentation and agent memory. Review the preview before any deletion or batch rewrite, keep the work in a git repository when possible, and be cautious if your agent memory or global guidance files contain private cross-project notes.

SkillSpector

Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The rollback guidance recommends `git restore .` or `git checkout -- <file>`, which can discard local uncommitted changes across the working tree. In an agent skill, this is risky because it may be followed automatically or semi-automatically during error recovery, causing unintended data loss without any requirement to preview, confirm scope, or ensure backups.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill explicitly instructs the agent to enumerate and read agent memory files and global configuration files, including locations under the user's home directory, without requiring user confirmation, scoping, or a warning about possible sensitive cross-project data exposure. In this context, that can cause over-collection of unrelated secrets, personal notes, or data from other repositories beyond the project being cleaned up.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill mandates actual file edits, file creation, and deletion of 'obsolete' files, but provides no safeguard requiring confirmation for destructive actions or any dry-run/review step. Because the protocol also frames deletion as a normal part of cleanup, an agent could remove files or memory content incorrectly, causing data loss or damaging project documentation state.

Session Persistence

Medium
Category
Rogue Agent
Content
### 第三步：实际修改（用工具，不只是描述）

你必须**真的用 Edit 修改现有文件、用 Write 创建新文件、用删除命令清理废弃文件**。"我会怎么改"的描述不算完成。

**顺序建议**：先改 docs/（改错影响外部）→ 再改 CLAUDE.md/AGENTS.md → 最后理记忆。先动外部优先级最高的，即使中途被打断，读者看到的也是对齐的最新状态。
Confidence
88% confidence
Finding
The protocol directs the agent to persist changes to docs, agent guidance files, and memory as part of routine execution, including creating and deleting files, without explicit constraints on what may be written into persistent stores. This creates a session-persistence risk: transient, unverified, or over-broad conversation content can be committed into durable memory/config artifacts and later influence future sessions or other projects.

Static analysis

No suspicious patterns detected.