ClawHub
Back to skill

Security audit

安全加密工具-免费版

Security checks for vulnerabilities and agentic risk

Overview

This encryption skill mostly matches its purpose, but it includes broad automatic execution language and optional irreversible file deletion without strong guardrails.

Review before installing. Use it only for explicit encrypt/decrypt tasks, confirm every input and output path, keep backups, and do not run shred/srm or overwrite originals until you have verified that the encrypted file can be decrypted successfully. Avoid putting real passphrases into reusable scripts or prompts unless you understand how your agent handles them.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The skill presents inconsistent trust and execution boundaries: it claims local-only operation and no API key requirement, yet also declares an LLM API as required. This can mislead users and host agents about where data may flow, increasing the chance that sensitive file names, passphrases, or content are exposed to remote model services unexpectedly.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The trigger condition is overly broad, covering unrelated contexts such as security detection, compliance audit, vulnerability scanning, and encryption protection. In an agent environment with exec/write permissions, vague activation criteria can cause the skill to run in contexts where users did not intend file encryption or related shell commands, increasing risk of unsafe or destructive actions.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The natural-language invocation model says users can trigger the scenario through general language and the tool will automatically execute corresponding operations, but it does not define strict constraints or confirmation steps. In an agent setting, this ambiguity can lead to unintended command execution on the wrong files or with unsafe defaults.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill includes destructive deletion commands such as shred/srm immediately after encryption without a prominent warning about irreversible data loss or prerequisite verification. If executed prematurely or on the wrong path, users can permanently lose original files before confirming the encrypted output is valid and recoverable.

Static analysis

No suspicious patterns detected.